An electromyography-based eye fatigue detection method, device, medium and program product

Abstract

The embodiment of the specification provides an eye fatigue detection method, device, medium and program product based on electromyographic detection, and relates to the field of intelligent medical treatment. The method comprises the following steps: receiving an original electromyographic signal of an orbicularis oculi muscle collected by an electromyographic sensor attached to a skin surface around eyes of a user; pre-processing the original electromyographic signal to obtain an electromyographic envelope signal, and identifying and marking a plurality of blink events from the electromyographic envelope signal based on a dynamic threshold method; extracting a plurality of multi-dimensional features for representing an eye fatigue state in the plurality of blink events, the multi-dimensional features comprising a frequency dimension feature, a force dimension feature and a fatigue dimension feature; calculating a blink frequency score according to the frequency dimension feature, calculating a blink force score according to the force dimension feature, and calculating a muscle fatigue score according to the fatigue dimension feature; and weighting and fusing the blink frequency score, the blink force score and the muscle fatigue score to obtain a comprehensive visual fatigue score.

Description

Technical Field

[0001] This invention relates to the field of intelligent medical care, and more specifically, to a method, device, medium, and program product for detecting eye fatigue based on electromyography (EMG). Background Technology

[0002] Visual fatigue is an increasingly prevalent health problem in modern society. With the widespread use of electronic devices and the increase in prolonged close-range visual activities, the global incidence of visual fatigue continues to rise. Visual fatigue not only affects visual comfort and work efficiency, but long-term accumulation can also lead to eye diseases such as dry eye syndrome and accommodative dysfunction, and even irreversible true myopia. Therefore, accurate, real-time, and continuous monitoring of visual fatigue has significant clinical importance and application value.

[0003] Currently, the assessment of visual fatigue mainly relies on subjective questionnaires (such as the OSDI scale) and objective clinical examinations (such as tear film breakup time and accommodation amplitude measurement). However, these methods have significant limitations: subjective questionnaires depend on the user's memory and perception, their accuracy is greatly affected by individual differences, and they cannot achieve real-time monitoring; clinical examinations need to be performed by professionals in specialized medical institutions, the equipment is expensive and time-consuming, and cannot meet the needs of daily continuous monitoring.

[0004] To achieve automated monitoring, existing technologies mainly include the following approaches: The first type is visual image-based detection technology. This type of technology acquires images of the user's face or eyes through a camera and uses computer vision algorithms to analyze indicators such as eyelid opening and closing, and blinking frequency. Although this technology is relatively intuitive, it has significant drawbacks: First, it involves facial image acquisition, posing a serious risk of user privacy leakage and limiting its application in private scenarios such as offices and homes; second, it has extremely high requirements for ambient lighting conditions, and its detection performance significantly decreases or even fails when the user is in dim light, backlight, or wearing glasses or sunglasses; in addition, factors such as facial occlusion and head movements can also seriously affect the accuracy of detection, and it requires high computing power, resulting in high device power consumption. The second type is fatigue detection technology based on multi-sensor fusion. For example, Chinese patent (CN109620221A Fatigue Reminder Method, Device, Smart Glasses and Medium Based on Smart Glasses) proposes to integrate electromyography sensors, vibration sensors and infrared cameras into smart glasses to comprehensively determine fatigue by combining multiple signals. While this method improves robustness, its hardware system is complex, requiring the integration of multiple different types of sensors, resulting in large, heavy, and costly devices. Simultaneously, the simultaneous operation of multiple sensors leads to a surge in power consumption and short battery life. More importantly, the data synchronization and fusion algorithms between different sensors are complex and suffer from data redundancy, hindering product lightweighting and widespread adoption. The third category is detection technology based on electromyography (EMG) signals. Existing technologies have already utilized EMG signals for blink detection or muscle state assessment. For example, research literature (Batulin et al., 2023) proposed a blink detection algorithm based on real-time assessment of the spectral power of orbicularis oculi muscle EMG signals, primarily for rehabilitation training of facial paralysis patients.

[0005] However, existing EMG-based technologies still have the following shortcomings when applied to visual fatigue detection: 1. The assessment dimensions are singular and lack comprehensiveness: For example, while comparative document 1 (202511326294.1-Mechanical EMG Signal Remote Monitoring Method and System) discloses the idea of ​​calculating fatigue index and frequency offset based on electromyographic signals, its core focus is on determining the "pathological state" (normal or abnormal) through a single-dimensional indicator (or a simple linear combination), without constructing a multi-dimensional comprehensive assessment system for the specific physiological state of "visual fatigue". Visual fatigue is a complex physiological process, and relying solely on a single fatigue index or frequency offset is insufficient to comprehensively reflect the real-time load state of the eye muscles.

[0006] 2. Lack of personalized calibration mechanism: The physiological characteristics of the eye orbicularis muscle vary greatly among different users (such as skin thickness, muscle strength, and basic blinking habits). Existing technologies mostly use uniform fixed thresholds or general models for analysis, without considering the significant impact of individual differences on EMG signal amplitude, duration, and other characteristics. This leads to poor consistency of test results among different users and a high likelihood of false positives or false negatives. Summary of the Invention

[0007] This invention aims to address at least one of the technical problems existing in the prior art. To this end, this invention provides a method, device, medium, and program product for detecting eye fatigue based on electromyography (EMG). The method of this invention, through a single periocular EMG sensor, can eliminate individual differences through personalized calibration and comprehensively quantify the degree of eye fatigue from three dimensions: frequency, intensity, and muscle fatigue. This solves problems such as privacy leaks, environmental limitations, hardware complexity, and inaccurate assessments in the prior art.

[0008] The first aspect of this application discloses a method for detecting eye fatigue based on electromyography (EMG), including: It receives raw electromyographic signals of the orbicularis oculi muscle acquired by an electromyographic sensor attached to the surface of the skin around the user's eyes; The original electromyographic signals are preprocessed to obtain electromyographic envelope signals, and multiple blinking events are identified and labeled from the electromyographic envelope signals based on the dynamic threshold method. Extract multi-dimensional features from the multiple blinking events to characterize eye fatigue, including frequency dimension features, intensity dimension features, and fatigue dimension features; The blink frequency score is calculated based on the frequency dimension feature, the blink intensity score is calculated based on the intensity dimension feature, and the muscle fatigue score is calculated based on the fatigue dimension feature. The blink frequency score, blink intensity score, and muscle fatigue score are weighted and fused to obtain a comprehensive visual fatigue score.

[0009] The second aspect of this application discloses an eye fatigue detection system based on electromyography (EMG) detection, used to implement the method disclosed in the first aspect of this application. The system includes: The data acquisition module is used to acquire raw electromyographic signals through an electromyography sensor; The personalized calibration module is used to perform personalized calibration steps and establish and store the user's personalized baseline parameters. The signal processing and event detection module is used to preprocess the raw electromyographic signals and identify blinking events; The multi-dimensional feature extraction and analysis module is used to extract frequency, intensity, and fatigue dimensions from the blinking event, and to call the personalized baseline parameters to calculate the corresponding dimension score and comprehensive visual fatigue score. The output module is used to output the comprehensive visual fatigue score and fatigue status information.

[0010] Optionally, the personalized calibration step includes: before monitoring begins, acquiring electromyographic signals corresponding to multiple calibration blink events generated by the user in a relaxed state, and establishing personalized baseline parameters for the user based on the calibration blink events, wherein the personalized baseline parameters include at least amplitude baseline, duration baseline, and area baseline.

[0011] A third aspect of this application discloses a computer device, comprising: a memory and a processor; the memory is used to store a computer program; the processor executes the computer program to implement the steps of the above-described method.

[0012] The fourth aspect of this application discloses a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the above-described method.

[0013] The fifth aspect of this application discloses a computer program product, including a computer program that, when executed by a processor, implements the steps of the above-described method.

[0014] This application has the following beneficial effects: 1. This application creatively proposes to utilize only a single periocular electromyography (EMG) sensor to deeply mine three major categories of features: frequency (blink frequency), intensity (relative amplitude, relative duration, relative integral area, and intensity consistency), and fatigue (median EMG frequency). The frequency dimension can sensitively capture reduced blink frequency and rhythmic disturbances caused by visual fatigue; the intensity dimension can quantitatively assess decreased contractile ability and reduced control stability of the orbicularis oculi muscle (e.g., incomplete blinking, excessive force); and the fatigue dimension can directly reflect the leftward shift of the spectrum caused by muscle metabolic fatigue from a frequency domain perspective. Furthermore, weighted fusion of the scores from these three dimensions allows for mutual supplementation and verification from different physiological mechanism levels, effectively avoiding misjudgments caused by momentary user movements or individual physical characteristics due to a single indicator.

[0015] 2. This application addresses the shortcomings of existing technologies where universal thresholds cannot adapt to the physiological differences of different users. It introduces a personalized calibration step before monitoring, effectively eliminating the impact of individual differences on the detection results. By guiding the user to blink naturally multiple times, user-specific baseline parameters are established, including amplitude baseline, duration baseline, and area baseline. In subsequent real-time monitoring, all force-related features are converted into relative values ​​relative to this user baseline (e.g., relative amplitude = current amplitude / baseline amplitude). This standardized processing mechanism allows the detection process to focus on the user's own changing trends rather than absolute voltage or time values, resulting in more accurate detection results.

[0016] 3. This application only requires a single electromyography (EMG) sensor attached to the skin surface around the eyes to complete all detection tasks. It does not require other hardware such as cameras, infrared sensors or vibration sensors, making it lightweight and not involving any facial image acquisition, thus eliminating the risk of privacy leakage. Furthermore, the EMG signals acquired in this application are directly derived from muscle electrophysiological activity and are completely unaffected by ambient light (strong light, dim light), facial obstruction (glasses, sunglasses, masks) and head movements. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a schematic diagram of the method flow provided in the first aspect of the present invention; Figure 2 This is a schematic diagram of an eye fatigue detection system based on electromyography provided in the second aspect of the present invention; Figure 3 These are EMG diagrams of blinking with different parameters provided in the embodiments of the present invention; Figure 4 It is the output result of calculation and evaluation after inputting information in the prediction process provided in the embodiments of the present invention. Detailed Implementation

[0019] To enable those skilled in the art to better understand the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings.

[0020] In some of the processes described in the specification, claims, and accompanying drawings of this invention, multiple operations appearing in a specific order are included. However, it should be clearly understood that these operations may not be executed in the order they appear herein, or may be executed in parallel. The operation numbers, such as 101, 102, etc., are merely used to distinguish different operations and do not represent any execution order. Furthermore, these processes may include more or fewer operations, and these operations may be executed sequentially or in parallel. It should be noted that the descriptions such as "first," "second," etc., in this document are used to distinguish different messages, devices, modules, etc., and do not represent a sequential order, nor do they limit "first" and "second" to different types.

[0021] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0022] Figure 1 This is a schematic flowchart of an eye fatigue detection method based on electromyography (EMG) provided by an embodiment of the present invention. Specifically, the method includes the following steps: S101: Receives raw electromyographic signals of the orbicularis oculi muscle acquired by an electromyographic sensor attached to the surface of the skin around the user's eyes; the sampling rate during acquisition is ≥500Hz; In some embodiments, the terms “user,” “subject,” “test subject,” or “sample” as used herein refer to any animal (e.g., a mammal), including but not limited to humans, non-human primates, rodents, etc., which will become the recipient of a particular treatment. Generally, the terms “user” and “patient” are used interchangeably herein when referring to human subjects. Preferably, the user is a human.

[0023] In some more specific embodiments, raw surface electromyographic signals of the orbicularis oculi muscle are acquired via the following hardware configuration: Electrode location: Upper edge of the orbital portion of the orbicularis oculi muscle (approximately 1 cm from the eyelid margin); Reference electrode: Center of cheek (grounded). Electrode type: Ag / AgCl wet electrode or conductive rubber dry electrode. Sampling rate: 500-1000Hz Amplification gain: 1000-5000 times. Input impedance: >100MΩ Common-mode rejection ratio: >80dB.

[0024] S102: The original electromyographic signal is preprocessed to obtain the electromyographic envelope signal, and multiple blinking events are identified and labeled from the electromyographic envelope signal based on the dynamic threshold method; In some embodiments, the identification and labeling of multiple blink events from the electromyographic envelope signal based on a dynamic thresholding method includes: Calculate the initial threshold of the electromyographic envelope signal, where the prime initial threshold is the first preset quantile of the signal amplitude; A noise statistic is calculated by selecting a signal below the initial threshold, and a dynamic detection threshold is determined based on the noise statistic. The electromyographic envelope signal is scanned, and the interval where the signal exceeds the dynamic detection threshold and conforms to preset waveform characteristics is identified as a blinking event. For example... Figure 3 The image shown is a schematic diagram of blinking EMG, with signal diagrams for bandpass filter parameters of 1Hz-100Hz, 1Hz-50Hz, 1Hz-10Hz, and 5Hz-10Hz.

[0025] In some embodiments, the preprocessing procedure for electromyographic signals is as follows: 1. Bandpass filter: 20-400Hz filter to remove low-frequency motion artifacts (<20Hz) and high-frequency noise (>400Hz); 2. Power frequency notch filter: 50 / 60Hz second-order IIR notch filter to remove power line interference; 3. Full-wave rectification: Taking the absolute value, converting a bidirectional signal into a unidirectional one; 4. Smoothing: The envelope is obtained by 10Hz low-pass filtering; 5. Baseline correction: Subtract the median of the signal to remove DC offset.

[0026] In some embodiments, the specific process for identifying and labeling blink events from EMG envelope signals based on preprocessed envelope signals includes: 1. Dynamic threshold calculation: Threshold = Baseline noise mean + 3 × Baseline noise standard deviation Baseline noise = the portion of the signal that is below the initial threshold.

[0027] 2. Event detection rules: Starting point: The signal exceeds the threshold and continues to rise. Peak point: The signal reaches a local maximum value. End point: The signal falls back below the threshold and remains below it for >100ms.

[0028] 3. Validation: Minimum duration: >80ms (excluding electromyographic pulses). Maximum duration: <500ms (excluding other facial movements) Minimum amplitude: >5 times the baseline noise Minimum interval: >200ms (to avoid duplicate counting).

[0029] As an example, the specific process of blink event recognition and labeling after obtaining EMG is described exemplarily. The parameters involved are not limiting to the core methodology of this embodiment. The specific process is as follows: 1. Adaptive threshold calculation Objective: To dynamically determine the signal threshold and distinguish blinking activity from baseline noise.

[0030] (1) Initial threshold: 70th percentile (values ​​higher than 70% of the signal). (2) Calculate noise statistics (the portion below the initial threshold). (3) Ensure that the threshold is not lower than the 60th percentile. Algorithm principle: 70th percentile: An empirical value that ensures most signals are below the threshold. 3σ principle: Based on noise distribution, set a statistical significance threshold. Double protection: Avoiding false detections due to excessively low thresholds. The threshold can be initially determined after the first wear, and then dynamically adjusted based on the electrode potential signal.

[0031] 2. Sliding window detection algorithm Core logic: Scan the entire signal to find regions that exceed the threshold and have a reasonable shape.

[0032] Step 1: Find the starting point (the point where the signal begins to rise). Step 2: Locate the peak point (the highest point of the signal). Step 3: Find the end point (signal falls back and is below the threshold). Step 4: Calculate features and verify.

[0033] 3. Validity verification Ensure that what is detected is a genuine blink, and not noise or artifacts: Duration: 80-500ms (normal blinking is about 100-400ms) Amplitude: At least twice the threshold (to ensure the signal is sufficiently significant). Shape: Both rise and fall times are >10ms (ensuring a complete contraction-relaxation process).

[0034] S103: Extract multi-dimensional features from the multiple blinking events to characterize eye fatigue, the multi-dimensional features including frequency dimension features, intensity dimension features and fatigue dimension features; In some embodiments, the frequency dimension feature includes blink frequency, which is calculated from multiple blink events; In some embodiments, the force dimension features include relative amplitude, relative duration, and relative integral area. The relative amplitude, relative duration, and relative integral area are calculated based on the signal amplitude, duration, and integral area of ​​each of the multiple blink events, and are standardized in combination with preset user-personalized baseline parameters.

[0035] In some embodiments, the preset user-personalized baseline parameters include amplitude baseline, duration baseline, and area baseline; the personalized baseline parameters are determined by the following method: before the monitoring begins, acquire electromyographic signals corresponding to multiple calibration phase blink events generated by the user in a relaxed state, and calculate the personalized baseline parameters; In some specific embodiments, personalized baseline parameters are determined by a personalized calibration module, primarily during initial use or periodic execution, prior to normal monitoring. By establishing user-personalized baseline parameters, individual differences are eliminated. The specific process includes: Calibration process: 1. Guidance Phase: Guide the user to blink 10 times naturally and relaxed. 2. Data Acquisition: Record EMG signals and blink events during the calibration phase. 3. Baseline calculation: Average peak amplitude = the average peak amplitude of 10 calibrated blinks. Average duration = the average duration of 10 calibrated blinks. Average integral area = the average integral area of ​​10 calibrated blinks. Resting noise level = standard deviation of signal during the blink-free period.

[0036] 4. Parameter storage: Save baseline parameters as a user configuration file.

[0037] In some embodiments, the fatigue dimension feature includes the median frequency of electromyography (EMG), which is extracted by performing frequency domain analysis on the original EMG signal or EMG envelope signal.

[0038] In some embodiments, the force dimension feature is calculated using the following method: For at least one blink event, the ratio of its peak amplitude to the amplitude baseline is calculated as the relative amplitude, the ratio of its duration to the duration baseline is calculated as the relative duration, and the ratio of its integral area to the area baseline is calculated as the relative integral area.

[0039] In some embodiments, after blink event detection and before score calculation, feature extraction and standardization processing are also included to extract a standardized feature vector from the blink event.

[0040] S104: Calculate the blink frequency score based on the frequency dimension feature, calculate the blink intensity score based on the intensity dimension feature, and calculate the muscle fatigue score based on the fatigue dimension feature.

[0041] In some embodiments, calculating the blink intensity score based on the intensity dimension features includes: Calculate the coefficient of variation of the relative amplitudes of multiple blink events within a preset time period; The force consistency index is determined based on the coefficient of variation; the lower the coefficient of variation, the higher the force consistency index. The blink intensity score is calculated based on the relative amplitude, relative duration, relative integral area, and intensity consistency index.

[0042] In some embodiments, calculating the muscle fatigue score based on fatigue dimension features includes: Calculate the decrease ratio of the currently extracted median electromyography frequency relative to the preset benchmark median frequency; The muscle fatigue score is calculated based on the decrease ratio, where a larger decrease ratio indicates a higher degree of fatigue.

[0043] S105: The blink frequency score, blink intensity score and muscle fatigue score are weighted and fused to obtain a comprehensive visual fatigue score.

[0044] As an example, the specific process of weighted fusion of blink frequency score, blink intensity score, and muscle fatigue score to obtain a comprehensive visual fatigue score is described exemplarily. The parameters involved are not limiting to the core methodology of this embodiment. The specific process is as follows: 1. Frequency-related characteristics, objective: to assess blink frequency and its stability.

[0045] (1) Current blinking frequency: Calculate the number of blinks in the last 60 seconds (converted to blinks / minute). (2) Frequency stability principle: The coefficient of variation of the interval between the last 10 blinks. The smaller the value, the more stable the frequency.

[0046] 2. Force-related features, core idea: standardize the current blinking feature to an individual baseline.

[0047] (1) Relative feature calculation: Use the most recent 10 blinks (or all, if less than 10). (2) Feature extraction: Each blink event obtained from the microcontroller includes: Peak amplitude, duration, integral area—area under the envelope.

[0048] (3) Calculate the relative eigenvalues ​​(standardization) Core formula: Relative feature = Current value / Individual baseline value Take the 10 most recent blinks (or all blinks if fewer than 10) and calculate the ratio relative to the baseline.

[0049] (4) Calculate the mean characteristics. (5) Consistency in calculation strength, Calculate the coefficient of variation of the relative amplitude, then subtract that value from 1. Coefficient of variation (CV) = Standard deviation / Mean Low CV (close to 0) → High consistency → Strength consistency close to 1. High CV (close to 0.5) → Low consistency → Strength consistency close to 0.5. Numerical examples: Assume the individual's baseline parameters are: amplitude_baseline: 120 μV duration_baseline: 180 ms area_baseline: 20000 μV·ms Three blinks were detected so far:

[0050] Calculation process: Relative amplitude: rel_amplitudes = [100 / 120, 110 / 120, 90 / 120] = [0.833, 0.917, 0.750], relative_amplitude = np.mean(rel_amplitudes) = 0.833.

[0051] Relative duration: rel_durations = [150 / 180, 170 / 180, 140 / 180] = [0.833, 0.944, 0.778], relative_duration = np.mean(rel_durations) = 0.852.

[0052] Relative area: rel_areas = [16000 / 20000, 18500 / 20000, 14000 / 20000] = [0.800, 0.925,0.700], relative_area = np.mean(rel_areas) = ​​0.808.

[0053] Consistency of force: The standard deviation of the amplitude is approximately 0.084, and the mean is 0.833. Coefficient of variation = 0.084 / 0.833 ≈ 0.101 strength_consistency = 1 - 0.101 = 0.899.

[0054] Physiological significance explanation

[0055] Typical changes during eye strain: A decrease in amplitude (<0.8): Muscle fatigue leads to weakened contractile force. Decreased area (<0.8): Overall muscle activity decreased. Decreased consistency (<0.8): Neuromuscular control is unstable.

[0056] 3. Fatigue-related characteristics, function: calculate signal complexity, and serve as an auxiliary indicator of fatigue.

[0057] Principle: The coefficient of variation of the signal rate of change is used to approximate the complex approximate entropy calculation.

[0058] Calculate the signal difference (rate of change) and use the coefficient of variation of the absolute value of the rate of change as a complexity index.

[0059] 4. Frequency Domain Analysis Conversion tool: Fourier transform Key frequency domain indicators (used for fatigue assessment) 1) Median frequency: The frequency point that divides the total area of ​​the power spectrum into two halves.

[0060] Fatigue symptoms: The median frequency decreases with muscle fatigue (typically by 10-30%).

[0061] 2) Average frequency: The weighted average frequency of the power spectrum.

[0062] 3) Power spectral ratio (high frequency / low frequency): The ratio decreases during fatigue.

[0063] It is necessary to explain the physiological mechanism by which fatigue causes a leftward shift in the spectrum: Changes in motor unit recruitment: During fatigue, fast-twitch fibers (high-frequency components) gradually become inactive, while slow-twitch fibers (low-frequency components) continue to work, resulting in an overall shift in the frequency spectrum towards lower frequencies. Decreased nerve conduction velocity: When muscles are fatigued, the nerve impulse conduction velocity slows down, the action potential duration is prolonged, and low-frequency components increase. Increased synchronization: In fatigued muscles, motor units discharge more synchronously, reducing high-frequency oscillations and increasing low-frequency components.

[0064] The relationship between complexity and fatigue: High complexity: Healthy muscle signals are highly variable. Low complexity: Fatigue muscle signals become regular and simple.

[0065] Transfer the above sets of data into the scoring module. (1) Calculate the degree of deviation of each item from 1.0 (2) Weighted composite (amplitude weight 40%, duration weight 30%, area weight 30%) (3) Convert to a 0-100 rating.

[0066] In some embodiments, the method further includes, after the score calculation is completed, classifying the risk based on the score and generating output, such as... Figure 4 As shown, this is the output result of calculation and evaluation after inputting information during the prediction process. The output mainly includes: 1. Quantitative data: Blinking frequency: X times / minute (normal range: 12-18). Frequency rating: X / 100 Strength rating: Y / 100 Fatigue rating: Z / 100 Overall score: T / 100.

[0067] 2. Risk Information: Risk level: Low / Medium / High risk Key abnormalities include: "blinking frequency reduced to 8 times / minute". Trend analysis: Compare with historical data (improvement / deterioration).

[0068] 3. Visual output: Real-time trend chart (rating changes over the past 30 minutes). Risk heat map (fatigue level at different time periods). Radar chart (comparison of various indicators).

[0069] 4. Intervention Recommendations: Low risk: Maintain good eye habits. Moderate risk: It is recommended to rest according to the 20-20-20 principle. High risk: Rest immediately and practice blinking.

[0070] In some embodiments, the auxiliary prediction results output based on the comprehensive visual fatigue score include, but are not limited to, paper or electronic reports. These results are obtained by intelligent machines based on the relevant data of the subjects and are only for reference by medical staff, and are not considered as the final diagnosis results of the subjects.

[0071] In some embodiments, the threshold mentioned above is obtained through training with training set samples. It can be a specific threshold or an interval range. The specific form is not specifically limited in this embodiment.

[0072] For ease of understanding, the training and application processes of the method involved in this embodiment are as follows: Training process (parameter optimization phase): Objective: To determine the optimal weight parameters and classification threshold in the algorithm.

[0073] Data requirements: Simultaneous acquisition: EMG signal + gold standard tag, Gold standard: OSDI questionnaire score, tear film breakup time (TBUT), and subjective fatigue score. Sample size: at least 30 participants (15 in the normal group + 15 in the eye fatigue group). Data balance: representativeness of different ages, genders, and eye habits.

[0074] Training process: Data collection → Data preprocessing → Feature extraction → Parameter initialization → Model training → Performance evaluation → Parameter saving.

[0075] Application process (real-time monitoring phase): Objective: To perform real-time evaluation of new users using trained parameters. Workflow: Device wearing → Personalized calibration → Real-time data acquisition → Preprocessing → Blink detection → Feature extraction → Score calculation → Risk classification → Real-time feedback.

[0076] A third aspect of this application discloses a computer device, which may include: one or more processors and one or more memories; wherein the memories store computer-readable code that, when run by the one or more processors, can perform the methods described above.

[0077] The processor in this embodiment can be an integrated circuit chip with signal processing capabilities. The processor can be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), an off-the-shelf programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. It can implement or execute the methods, operations, and logic block diagrams disclosed in this embodiment. The general-purpose processor can be a microprocessor or any conventional processor, and can be based on an x86 or ARM architecture.

[0078] In general, the various exemplary embodiments of this disclosure can be implemented in hardware or dedicated circuitry, software, firmware, logic, or any combination thereof. Some aspects can be implemented in hardware, while others can be implemented in firmware or software that can be executed by a controller, microprocessor, or other computing device. When aspects of embodiments of this disclosure are illustrated or described as block diagrams, flowcharts, or using some other graphical representation, it will be understood that the blocks, apparatuses, systems, techniques, or methods described herein can be implemented as non-limiting examples in hardware, software, firmware, dedicated circuitry or logic, general-purpose hardware or controllers or other computing devices, or some combination thereof.

[0079] For example, the methods or apparatus according to embodiments of this disclosure can also be implemented using the architecture of a computing device. A computing device may include a bus, one or more CPUs, read-only memory (ROM), random access memory (RAM), a communication port connected to a network, input / output components, a hard disk, etc. Storage devices in the computing device, such as ROM or hard disk, may store various data or files used for processing and / or communication of the methods provided in this disclosure, as well as program instructions executed by the CPU. The computing device may also include a user interface. Of course, the architecture described above is merely exemplary, and one or more components of the computing device described above may be omitted as needed when implementing different devices.

[0080] This invention also provides a computer-readable storage medium storing computer-readable instructions. When the computer-readable instructions are executed by a processor, the methods disclosed in this embodiment can be performed. The computer-readable storage medium in this embodiment can be volatile memory or non-volatile memory, or may include both volatile and non-volatile memory. Non-volatile memory can be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), or flash memory. Volatile memory can be random access memory (RAM), which is used as an external cache. By way of example, but not limitation, many forms of RAM are available, such as static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDRSDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous interconnected dynamic random access memory (SLDRAM), and direct memory bus random access memory (DRRAM). It should be noted that the memory of the methods described herein is intended to include, but is not limited to, these and any other suitable types of memory.

[0081] This disclosure also provides a computer program product or system, including a computer program that, when executed by a processor, implements the steps of the above-described method, such as... Figure 2 As shown, the computer program product or computer program is used to execute an eye fatigue detection system based on electromyography (EMG) detection, the system comprising: Data acquisition module 201 is used to acquire raw electromyographic signals through an electromyographic sensor; Personalized calibration module 202 is used to perform personalized calibration steps, establish and store the user's personalized baseline parameters; The signal processing and event detection module 203 is used to preprocess the raw electromyographic signals and identify blinking events; The multi-dimensional feature extraction and analysis module 204 is used to extract frequency, intensity and fatigue dimension features from the blinking event, and call the personalized baseline parameters to calculate the corresponding dimension score and comprehensive visual fatigue score; The output module 205 is used to output the comprehensive visual fatigue score and fatigue status information.

[0082] The exemplary embodiments of this disclosure described in detail above are merely illustrative and not restrictive. Those skilled in the art will understand that various modifications and combinations can be made to these embodiments or their features without departing from the principles and spirit of this disclosure, and such modifications should fall within the scope of this disclosure.

Claims

1. A method for detecting eye fatigue based on electromyography (EMG), characterized in that, include: It receives raw electromyographic signals of the orbicularis oculi muscle acquired by an electromyographic sensor attached to the surface of the skin around the user's eyes; The original electromyographic signals are preprocessed to obtain electromyographic envelope signals, and multiple blinking events are identified and labeled from the electromyographic envelope signals based on the dynamic threshold method. Multi-dimensional features for characterizing eye fatigue are extracted from the multiple blink events. These multi-dimensional features include frequency, intensity, and fatigue dimensions. The frequency dimension features include blink frequency, calculated from multiple blink events. The intensity dimension features include relative amplitude, relative duration, and relative integral area. These are calculated by standardizing the signal amplitude, duration, and integral area of ​​each blink event using preset user-personalized baseline parameters. The preset user-personalized baseline parameters include amplitude baseline, duration baseline, and area baseline. The intensity dimension features are calculated as follows: for at least one blink event, the ratio of its peak amplitude to the amplitude baseline is calculated as the relative amplitude; the ratio of its duration to the duration baseline is calculated as the relative duration; and the ratio of its integral area to the area baseline is calculated as the relative integral area. The fatigue dimension features include the median frequency of electromyography (EMG), extracted by performing frequency domain analysis on the original EMG signal or EMG envelope signal. A blink frequency score is calculated based on the frequency dimension feature, a blink intensity score is calculated based on the intensity dimension feature, and a muscle fatigue score is calculated based on the fatigue dimension feature. Specifically, calculating the blink intensity score based on the intensity dimension feature includes: calculating the coefficient of variation of the relative amplitudes of multiple blink events within a preset time period; determining an intensity consistency index based on the coefficient of variation; and calculating the blink intensity score based on the relative amplitude, relative duration, relative integral area, and intensity consistency index. Calculating the muscle fatigue score based on the fatigue dimension feature includes: calculating the decrease ratio of the currently extracted median electromyographic frequency relative to a preset benchmark median frequency; and calculating the muscle fatigue score based on the decrease ratio. The blink frequency score, blink intensity score, and muscle fatigue score are weighted and fused to obtain a comprehensive visual fatigue score.

2. The method for detecting eye fatigue based on electromyography according to claim 1, characterized in that, The personalized baseline parameters are determined by acquiring electromyographic signals corresponding to multiple calibration phase blink events generated by the user in a relaxed state before the monitoring begins, and calculating the personalized baseline parameters.

3. The method for detecting eye fatigue based on electromyography according to claim 1, characterized in that, The lower the coefficient of variation, the higher the strength consistency index.

4. The method for detecting eye fatigue based on electromyography according to claim 1, characterized in that, The higher the decrease ratio, the higher the degree of fatigue indicated by the muscle fatigue score.

5. The method for detecting eye fatigue based on electromyography according to claim 1, characterized in that, The method of identifying and labeling multiple blink events from the electromyographic envelope signal based on a dynamic threshold method includes: Calculate the initial threshold of the electromyographic envelope signal, where the initial threshold is the first preset quantile of the signal amplitude; A noise statistic is calculated by selecting a signal below the initial threshold, and a dynamic detection threshold is determined based on the noise statistic. The electromyographic envelope signal is scanned, and the interval where the signal exceeds the dynamic detection threshold and conforms to the preset waveform characteristics is identified as a blinking event.

6. An eye fatigue detection system based on electromyography (EMG), characterized in that, The system for implementing the method according to any one of claims 1-5 comprises: The data acquisition module is used to acquire raw electromyographic signals through an electromyography sensor; The personalized calibration module is used to perform personalized calibration steps and establish and store the user's personalized baseline parameters. The signal processing and event detection module is used to preprocess the raw electromyographic signals and identify blinking events; The multi-dimensional feature extraction and analysis module is used to extract frequency, intensity, and fatigue dimensions from the blinking event, and to call the personalized baseline parameters to calculate the corresponding dimension score and comprehensive visual fatigue score. The output module is used to output the comprehensive visual fatigue score and fatigue status information; The personalized calibration step includes: before the monitoring begins, acquiring electromyographic signals corresponding to multiple calibration phase blink events generated by the user in a relaxed state, and establishing personalized baseline parameters for the user based on the calibration phase blink events. The personalized baseline parameters include at least amplitude baseline, duration baseline, and area baseline.

7. A computer device, characterized in that, The device includes: a memory and a processor; the memory is used to store a computer program; the processor executes the computer program to implement the steps of the method according to any one of claims 1-5.

8. A computer-readable storage medium, characterized in that, It stores a computer program that, when executed by a processor, implements the steps of the method as described in any one of claims 1-5.

9. A computer program product, comprising a computer program, characterized in that, When executed by a processor, the computer program implements the steps of the method described in any one of claims 1-5.

Images