Myopia early warning method and system based on children and teenagers

Abstract

The invention discloses a myopia early warning method and system based on children and teenagers, and relates to the technical field of health monitoring, and the method comprises the steps: collecting bio-electricity signal data and environment sensor data in real time; through combination of a convolutional neural network and a recurrent neural network, feature extraction is carried out on the collected bio-electricity signal data, a fatigue signal and a fatigue degree are identified, the current visual fatigue state of children and adolescents is evaluated, and a fatigue score is output; and monitoring the fatigue score of the children and adolescents in real time, and when the fatigue score exceeds a myopia early warning threshold value, automatically triggering an early warning signal and generating a prompt and eyesight protection suggestion for the current myopia risk of the children and adolescents. According to the invention, through combination with environmental data, when fatigue scores are monitored in real time, environmental conditions can be automatically adjusted in combination with an intelligent device, the learning and living environment of children is optimized, and the occurrence risk of myopia is further reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of health monitoring, and in particular to a method and system for early warning of myopia in children and adolescents. Background Art

[0002] With the rapid development of artificial intelligence and sensing technology, health monitoring systems based on bioelectric signals and environmental sensor data have made significant progress in many fields. In the field of visual health of children and adolescents, researchers have tried to use various technical means, such as visual fatigue monitoring and myopia warning, to improve the vision of adolescents. In recent years, deep learning algorithms, especially the combination of convolutional neural networks (CNN) and recurrent neural networks (RNN), have been widely used in medical signal processing and health management. By collecting bioelectric signals, behavioral data and environmental data in real time, combined with intelligent algorithms, early identification and intervention of visual fatigue can be achieved to prevent further deterioration of myopia problems.

[0003] However, existing technologies still have limitations in accurately assessing the visual fatigue and health status of children and adolescents. Most existing systems rely mainly on static thresholds for health monitoring and lack the ability to be personalized and dynamically adjusted. In particular, there are technical deficiencies in the relationship between fatigue assessment and myopia warning and environmental adaptation. In addition, the warning mechanism of the existing system is too simple and cannot adjust the monitoring strategy in real time according to environmental changes and individual differences, which may lead to misjudgment or miss the best time for intervention. Therefore, how to integrate multiple data sources and adopt more flexible and personalized algorithms for real-time warning has become a key issue that needs to be solved urgently. Summary of the invention

[0004] In view of the above existing problems, the present invention is proposed.

[0005] Therefore, the present invention provides a method for early warning of myopia in children and adolescents to solve the problem that the prior art lacks a personalized, real-time, and dynamically adjusted early warning system in visual health monitoring of children and adolescents.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0007] In a first aspect, the present invention provides a method for early warning of myopia in children and adolescents, which includes real-time collection of bioelectric signal data and environmental sensor data;

[0008] By combining convolutional neural networks with recurrent neural networks, the collected bioelectric signal data is feature extracted to identify fatigue signals and fatigue levels, and the current visual fatigue status of children and adolescents is evaluated, and a fatigue score is output;

[0009] Based on fatigue scores and personal health records of children and adolescents, combined with reinforcement learning algorithms, a personalized dynamic warning model is built according to the physiological and behavioral characteristics of children in bioelectric signal data, and a myopia warning threshold is set;

[0010] Analyze current environmental conditions by combining environmental sensor data and myopia warning thresholds, and adjust environmental conditions in real time through associated smart devices;

[0011] Monitor the fatigue scores of children and adolescents in real time. When the fatigue score exceeds the myopia warning threshold, it will automatically trigger a warning signal and generate reminders and vision protection recommendations for the current myopia risk of children and adolescents.

[0012] As a preferred solution of the method for early warning of myopia in children and adolescents described in the present invention, wherein: the bioelectric signal data includes electroencephalogram (EEG), electrocardiogram (ECG) and electromyogram (EMG);

[0013] The environmental sensor data includes temperature and humidity sensors, gas sensors and light sensors.

[0014] As a preferred solution of the method for early warning of myopia in children and adolescents described in the present invention, wherein: the convolutional neural network is combined with the recurrent neural network to extract features from the collected bioelectric signal data to identify fatigue signals and fatigue levels. The specific steps are as follows:

[0015] Remove noise and standardize bioelectric signal data to retain effective bioelectric signals;

[0016] Extract the convolutional neural network spatial features from effective bioelectric signals, and use wavelet transform to enhance the high-frequency and low-frequency features of bioelectric signals;

[0017] Input high-frequency features and low-frequency features into the LSTM network, learn the temporal features of high-frequency features and low-frequency features, then fuse the temporal features through the fully connected layer and output the fused features;

[0018] The fused features are input into the fully connected network to identify the fatigue signal. SoftMax receives the fatigue signal and outputs the classification probability of the fatigue degree.

[0019] As a preferred solution of the method for early warning of myopia in children and adolescents according to the present invention, the steps of evaluating the current visual fatigue state of children and adolescents and outputting fatigue scores are as follows:

[0020] Extract the probability value of each fatigue level from the classification probability output by the fully connected network;

[0021] Set corresponding scoring intervals of mild fatigue, moderate fatigue and severe fatigue for each fatigue level;

[0022] Using the classification probability value of each fatigue level, the corresponding scoring intervals are weighted averaged to obtain the final fatigue score.

[0023] As a preferred solution of the method for early warning of myopia in children and adolescents described in the present invention, wherein: based on fatigue scores and personal health records of children and adolescents, combined with a reinforcement learning algorithm, a personalized dynamic early warning model is constructed according to the physiological characteristics and behavioral characteristics of children in bioelectric signal data, and a myopia early warning threshold is set. The specific steps are as follows:

[0024] Collect the daily eye use time and sitting posture behavior characteristics of children and adolescents in real time;

[0025] Combine the bioelectric signal data and behavioral characteristics of children and adolescents to build personal health records;

[0026] The data in the personal health record is input into the multimodal neural network for feature fusion to obtain a comprehensive personalized feature vector;

[0027] According to the personalized feature vector, bioelectric signal data and behavioral characteristics, a personalized dynamic early warning model is constructed, which is expressed as:

[0028]

[0029] F(ΔX(τ))=ΔX(τ) 2 ;

[0030] Among them, P(t) is the output of the personalized dynamic warning model at time t, n is the bioelectric signal feature, m is the number of behavioral features, X i (t) is the value of the i-th feature in the bioelectric signal data at time t, w i is the weight corresponding to the i-th feature, Y j (t) is the value of the jth feature in the behavior feature at time t, v j is the weight corresponding to the jth feature; ΔX(τ) is the change in the time interval, τ is the integral variable, dτ is the small change, λ is the attenuation parameter, and F(ΔX(τ)) is the time difference function;

[0031] Use reinforcement learning algorithms to train personalized dynamic warning models;

[0032] Based on the training results of the personalized dynamic warning model, the myopia warning threshold is set through a dynamic adjustment strategy.

[0033] As a preferred solution of the method for early warning of myopia in children and adolescents described in the present invention, the following specific steps are used to analyze the current environmental conditions in combination with the environmental sensor data and the myopia early warning threshold, and to control the environmental conditions in real time through the associated intelligent device:

[0034] Collect light, temperature, humidity and air quality data in real time through environmental sensors;

[0035] Analyze current environmental conditions by combining environmental sensor data and myopia warning thresholds;

[0036] The Q-learning algorithm is used to calculate and analyze the results and generate an optimized environmental adjustment strategy;

[0037] Based on the optimized environmental adjustment strategy, by associating intelligent devices, the device parameters are automatically adjusted and the environmental conditions are controlled in real time.

[0038] As a preferred solution of the method for early warning of myopia in children and adolescents according to the present invention, wherein: the fatigue score of children and adolescents is monitored in real time, and when the fatigue score exceeds the myopia early warning threshold, an early warning signal is automatically triggered and a prompt and vision protection suggestion for the current myopia risk of children and adolescents are generated. The specific steps are as follows:

[0039] Under environmental condition monitoring, a personalized dynamic early warning model is used to collect and analyze children and adolescents’ fatigue scores in real time;

[0040] When the fatigue score exceeds the myopia warning threshold, the warning signal is automatically triggered;

[0041] When the warning signal is triggered, the associated smart device issues a reminder about the current myopia risk to alert children and adolescents;

[0042] Input myopia risk reminders, fatigue scores, smart devices and bioelectric signal data feedback into a personalized dynamic early warning model to further optimize the early warning strategy;

[0043] Generate vision protection recommendations based on the optimized early warning strategy and real-time data.

[0044] In a second aspect, the present invention provides a myopia warning system for children and adolescents, including a data collection module, a feature extraction module, a model building module, an environment control module and a warning prompt module;

[0045] The data acquisition module is used to collect bioelectric signal data and environmental sensor data in real time;

[0046] The feature extraction module is used to extract features from the collected bioelectric signal data by combining a convolutional neural network with a recurrent neural network, identify fatigue signals and fatigue levels, and evaluate the current visual fatigue state of children and adolescents, and output a fatigue score;

[0047] The model building module is used to build a personalized dynamic warning model based on fatigue scores and personal health records of children and adolescents, combined with a reinforcement learning algorithm, according to the physiological and behavioral characteristics of children in bioelectric signal data, and set a myopia warning threshold;

[0048] The environmental control module is used to analyze the current environmental conditions in combination with the environmental sensor data and the myopia warning threshold, and to control the environmental conditions in real time by associating with the smart device;

[0049] The early warning prompt module is used to monitor the fatigue scores of children and adolescents in real time. When the fatigue score exceeds the myopia warning threshold, it automatically triggers a warning signal and generates prompts and vision protection suggestions for the current myopia risk of children and adolescents.

[0050] In a third aspect, the present invention provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program is executed by the processor, any step of the method for early warning of myopia in children and adolescents as described in the first aspect of the present invention is implemented.

[0051] In a fourth aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, any step of the method for early warning of myopia in children and adolescents as described in the first aspect of the present invention is implemented.

[0052] The beneficial effects of the present invention are as follows: the present invention proposes a personalized dynamic early warning model by combining bioelectric signals, environmental sensor data and deep learning algorithms, which can monitor the visual fatigue status of children and adolescents in real time and make intelligent adjustments. Compared with the traditional static threshold early warning, the present invention effectively improves the recognition accuracy of fatigue signals through the feature extraction method combining convolutional neural network (CNN) and recurrent neural network (RNN), and can dynamically adapt to the differences in individual physiological characteristics. On this basis, combined with the reinforcement learning algorithm, it can intelligently adjust the early warning model according to the child’s health record and real-time data, and provide personalized health intervention measures. In particular, the combination with environmental data can automatically adjust the environmental conditions in combination with intelligent devices when monitoring fatigue scores in real time, optimize children’s learning and living environment, and further reduce the risk of myopia. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use 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 ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0054] Figure 1 This is a flow chart of the myopia early warning method for children and adolescents in Example 1.

[0055] Figure 2 This is a module diagram of the myopia early warning system for children and adolescents in Example 1. DETAILED DESCRIPTION

[0056] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the accompanying drawings.

[0057] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0058] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive with other embodiments.

[0059] Example 1, reference Figure 1 and Figure 2 , which is the first embodiment of the present invention, and provides a method for early warning of myopia in children and adolescents, comprising the following steps:

[0060] S1. Collect bioelectric signal data and environmental sensor data in real time.

[0061] Furthermore, the bioelectric signal data includes electroencephalogram (EEG), electrocardiogram (ECG) and electromyogram (EMG); the environmental sensor data includes temperature and humidity sensors, gas sensors and light sensors.

[0062] Specifically, bioelectric signal data is collected through high-precision electroencephalogram (EEG), electrocardiogram (ECG) and electromyogram (EMG) sensors. These devices can monitor the physiological status of children and adolescents in real time, including brain electrical activity, heart rate and muscle tension, so as to accurately reflect their fatigue level; through Bluetooth and Wi-Fi wireless technology, bioelectric signal data is quickly transmitted to the data processing center for real-time analysis; at the same time, environmental sensor data is collected through temperature and humidity sensors, gas sensors and light sensors to obtain real-time information such as temperature, humidity, air quality and light intensity in the environment; all collected data are aggregated through wireless networks and transmitted to the cloud platform or local server for analysis and processing;

[0063] Preferably, fatigue scores and personalized early warning models provide comprehensive and real-time data support to ensure the dynamic and efficient nature of the program, conduct precise health monitoring and timely intervention based on real-time data, greatly improve the accuracy and timeliness of visual health management, and effectively reduce the risk of vision problems caused by fatigue or uncomfortable environment in children and adolescents.

[0064] S2. By combining convolutional neural networks with recurrent neural networks, the collected bioelectric signal data is feature extracted to identify fatigue signals and fatigue levels, and the current visual fatigue status of children and adolescents is evaluated to output a fatigue score.

[0065] Furthermore, the bioelectric signal data is subjected to noise removal and standardization processing to retain the effective bioelectric signal;

[0066] It should be noted that effective bioelectric signals refer to signals that can truly reflect the physiological state of the human body after removing noise and interference, including bioelectric signals such as electroencephalogram (EEG), electrocardiogram (ECG) and electromyogram (EMG), which can accurately show physiological changes such as fatigue and stress. Effective bioelectric signals retain their true physiological information after noise removal and are standardized for subsequent analysis and feature extraction.

[0067] Specifically, a bandpass filter is used to select a frequency range of interest in the signal and filter out noise above or below the range. For example, for EEG signals, the frequency range of 0.5-50 Hz is usually of interest, while noise with a frequency higher than 50 Hz (such as power frequency noise) or artifacts with a frequency lower than 0.5 Hz (such as eye movement artifacts) can be removed by bandpass filtering. The bandpass filter will effectively retain the useful frequency components in the signal while eliminating unnecessary noise.

[0068] Standardization is to convert data into a unified scale so that signals from different sources can be processed under the same standard. Bioelectric signal data usually have large differences at different times, environmental conditions and between individuals. Standardization can eliminate these differences, making subsequent feature extraction and pattern recognition more accurate and stable.

[0069] Use ICA to retain valid physiological signals. ICA is a signal separation technology that can extract independent source signals from multi-channel signals. In EEG or ECG signal processing, ICA can effectively separate EEG signals from noise sources such as eye movement artifacts and ECG artifacts, remove these interferences, and retain valid physiological signals.

[0070] Preferably, through the noise removal method, various electromagnetic interferences, artifacts and other irrelevant signals can be removed from the bioelectric signal, and the effective data that truly reflects the physiological state can be retained, which effectively improves the quality of the signal and provides an accurate basis for subsequent feature extraction and analysis; through standardization processing, the differences between different signal sources and devices are eliminated, so that the signals can be uniformly processed under the same standard, which not only improves the consistency between different signal channels, but also improves the algorithm's processing ability for various signal features, which is especially important for the subsequent processing combined with CNN, LSTM and other algorithms;

[0071] Extract the convolutional neural network spatial features from effective bioelectric signals, and use wavelet transform to enhance the high-frequency and low-frequency features of bioelectric signals;

[0072] Specifically, bioelectric signals (such as EEG, ECG, and EMG) have complex structures and multi-level features in spatial and temporal dimensions. The successful application of convolutional neural networks (CNNs) in the field of computer vision has shown that they can effectively extract local features in images and signals and capture spatial structural information through multi-layer convolution operations. Convolutional neural networks (CNNs) are applied to the spatial feature extraction of bioelectric signals.

[0073] The convolution layer uses multiple convolution kernels to locally filter the input bioelectric signals and extract local features. In bioelectric signal processing, each convolution kernel can learn the signal pattern of a specific frequency or spatial position. For example, in EEG signal processing, CNN can extract spatial distribution features from signals at different electrode positions, which is particularly important for understanding the relationship between signal sources (such as brain area activity).

[0074] CNN gradually extracts features from low to high layers through multi-layer convolution operations. The initial convolution kernel extracts low-level features, such as simple electrical signal waveform features, while the deep convolution kernel can extract more abstract features, such as spectrum structure, time series pattern, etc. This hierarchical learning enables CNN to capture complex spatial information in the signal, thereby more accurately reflecting the physical and physiological background of the signal.

[0075] Wavelet transform can decompose the signal into components of multiple scales (frequencies). Through wavelet decomposition, the high-frequency and low-frequency parts of the signal are effectively distinguished. The high-frequency part usually corresponds to faster changes (such as fluctuations in a short time), and the low-frequency part corresponds to the stable part of the signal (such as changes on a longer time scale). In EEG signals, high-frequency components usually represent rapid brain activity, while low-frequency components represent relatively slow EEG activity.

[0076] Finally, the signal is subjected to wavelet transform, which can enhance its high-frequency and low-frequency features. For example, high-frequency feature enhancement helps capture rapidly changing physiological phenomena (such as sudden brain wave changes or abnormal states), while low-frequency feature enhancement helps reflect long-term trends and laws (such as brain wave patterns in deep sleep stages). Through the processing of wavelet coefficients, these features can be effectively amplified to make them more prominent in subsequent analysis.

[0077] It should be noted that bioelectric signals contain different physiological information in different frequency ranges. Especially in EEG signals, low-frequency (such as delta waves, theta waves) and high-frequency (such as alpha waves, beta waves) signals often reflect different brain activity states (such as wakefulness, fatigue, relaxation, etc.). Wavelet transform is a multi-resolution analysis method that can decompose signals in two dimensions, time and frequency, and is a very effective tool in the field of signal processing.

[0078] Advantageously, wavelet transform can provide more detailed local time-frequency information than Fourier transform, and is particularly effective for analyzing complex bioelectric signals. By enhancing the high-frequency and low-frequency characteristics of bioelectric signals, it can effectively capture the signal patterns of rapid changes and long-term changes under fatigue conditions, thereby providing accurate data for subsequent fatigue identification;

[0079] Input high-frequency features and low-frequency features into the LSTM network, learn the temporal features of high-frequency features and low-frequency features, then fuse the temporal features through the fully connected layer and output the fused features;

[0080] Specifically, the LSTM network can learn the changes of signals over time through its memory units, identify long-term dependencies in the signals, and accurately capture the accumulation process of fatigue. LSTM can effectively avoid the gradient vanishing problem that may occur in traditional RNN when processing long sequence data, thereby better analyzing the timing information in bioelectric signals;

[0081] Preferably, by introducing LSTM, the changing trend of fatigue status can be identified more accurately, which provides stronger time series modeling capability for fatigue signal detection and improves the accuracy of response to continuous fatigue signals;

[0082] The fused features are input into the fully connected network to identify the fatigue signal. SoftMax receives the fatigue signal and outputs the classification probability of the fatigue degree.

[0083] Specifically, the input of the fully connected layer is the fused feature vector output by CNN and LSTM, which contains the information of the signal in spatial and temporal dimensions;

[0084] The fully connected layer assigns a weight to each feature input and combines all input features into a unified output vector by weighted summation;

[0085] The result after weighted summation will be transformed nonlinearly through activation functions (such as ReLU, Sigmoid, etc.), so that the network can learn complex signal features;

[0086] The output of the fully connected layer is a vector, each element of which represents the score of a specific category. Assuming that the network needs to output the classification results of three categories: mild fatigue, moderate fatigue, and severe fatigue, SoftMax receives a vector of length 3, which contains the scores of these three categories.

[0087] SoftMax calculates the exponential function for each category and then standardizes the values ​​so that the sum of the probabilities of all categories is 1. In this way, the score of each category is converted into a corresponding probability, reflecting the possibility that the signal belongs to that category.

[0088] It should be noted that the fully connected network (FCN) is a common layer structure in neural networks, usually used for comprehensive processing of signals or features. Fusion features refer to the features obtained by combining convolutional neural networks (CNN) and long short-term memory networks (LSTM). These features contain the spatial information and temporal information of bioelectric signals.

[0089] The SoftMax function is a commonly used activation function in multi-classification problems. It converts the output of the neural network into a probability distribution and is usually used in the output layer of the neural network to perform probability estimation on the classification task. In the present invention, the SoftMax function receives the fusion features output by the fully connected layer and converts them into the classification probability of the fatigue degree.

[0090] Extract the probability value of each fatigue level from the classification probability output by the fully connected network;

[0091] Set corresponding scoring intervals of mild fatigue, moderate fatigue and severe fatigue for each fatigue level;

[0092] Using the classification probability value of each fatigue level, the corresponding scoring intervals are weighted averaged to obtain the final fatigue score;

[0093] It should be noted that the classification probability value means that after the neural network is trained, for each category (here, fatigue level), the network will give a value, indicating the probability that the input signal belongs to a certain category. The classification probability value here measures the credibility of different fatigue levels;

[0094] Through the fatigue level, different fatigue states (such as mild, moderate, and severe) are obtained. The fatigue level reflects the individual fatigue state evaluated based on bioelectric signals and other related data (such as physiological characteristics, timing information, etc.);

[0095] Each fatigue level is assigned a specific scoring interval (e.g., mild fatigue corresponds to one scoring interval, severe fatigue corresponds to another scoring interval), which is set according to the critical criteria of the level in order to weight the final fatigue score according to the value of the classification probability;

[0096] Weighted average means that when calculating the final fatigue score, different weights are assigned to the probability values ​​of each fatigue level, and then the weighted average value is obtained. This can more accurately determine the final score based on the classification probability value and avoid errors caused by a single classification result;

[0097] Preferably, the classification probability values ​​output by the fully connected network are weighted averaged and combined with the scoring range of each fatigue level to obtain the final fatigue score, which solves several key problems in traditional fatigue assessment methods. First, it uses the classification probability output by the deep learning model rather than a single classification result, making fatigue assessment more flexible and accurate; secondly, it integrates the probability information of different fatigue levels by weighted averaging, avoiding the accumulation of classification errors and improving the comprehensive assessment ability of fatigue status.

[0098] S3. Based on fatigue scores and personal health records of children and adolescents, combined with reinforcement learning algorithms, a personalized dynamic warning model is constructed according to the physiological and behavioral characteristics of children in bioelectric signal data, and the myopia warning threshold is set.

[0099] Furthermore, the daily eye use time and sitting posture behavior characteristics of children and adolescents are collected in real time;

[0100] Specifically, use smart glasses, wearable devices (such as smart watches) or mobile phone applications to monitor children's visual activities. These devices use built-in light sensors, accelerometers and other sensors to detect and record children's visual activity time in real time. For example, the device can detect changes in screen brightness or the intensity of ambient light when wearing glasses to infer the length of time children use their eyes;

[0101] Use smart wearable devices (such as smart clothing, seat sensors, etc.) to monitor children's sitting posture in real time, and use accelerometers, gyroscopes, etc. to detect children's body posture, especially changes in the back, waist and spine;

[0102] It should be noted that behavioral characteristics refer to the activity data of children and adolescents in their daily lives, such as the time they use their eyes, their sitting posture, the amount of exercise they do, etc. These behavioral characteristics can affect their visual health and therefore need to be monitored and quantified in real time;

[0103] Combine the bioelectric signal data and behavioral characteristics of children and adolescents to build personal health records;

[0104] Specifically, first, the bioelectric signal data (such as electroencephalogram, electrocardiogram, electromyogram) and behavioral characteristic data (such as eye use time, sitting posture) of children and adolescents are collected in real time;

[0105] Secondly, clean and standardize the data to ensure its accuracy and consistency;

[0106] Then, the bioelectric signals and behavioral characteristics are fused through a multimodal neural network to extract personalized health feature vectors;

[0107] Finally, based on these integrated characteristics and combined with the individual differences of children, a personal health profile is constructed, and updated and optimized based on real-time data to ensure the accuracy of health monitoring and early warning.

[0108] The data in the personal health record is input into the multimodal neural network for feature fusion to obtain a comprehensive personalized feature vector;

[0109] It should be noted that fusing feature data from different sources (such as bioelectric signals, behavioral characteristics, etc.) can effectively extract individual health information in all aspects. This method can not only improve the accuracy of data processing, but also tap more potential correlations. Compared with traditional single data source processing methods, multimodal neural networks can achieve more accurate and comprehensive feature representation, thereby enhancing the accuracy of personalized health assessment.

[0110] According to the personalized feature vector, bioelectric signal data and behavioral characteristics, a personalized dynamic early warning model is constructed, which is expressed as:

[0111]

[0112] F(ΔX(τ))=ΔX(τ) 2 ;

[0113] Among them, P(t) is the output of the personalized dynamic warning model at time t, n is the bioelectric signal feature, m is the number of behavioral features, X i (t) is the value of the i-th feature in the bioelectric signal data at time t, w i is the weight corresponding to the i-th feature, Y j (t) is the value of the jth feature in the behavior feature at time t, v j is the weight corresponding to the jth feature; ΔX(τ) is the change in the time interval, τ is the integral variable, dτ is the small change, λ is an attenuation parameter that controls the influence range of signal changes, and F(ΔX(τ)) is the time difference function that measures the change of the signal in the time interval [0, t];

[0114] Specifically, P(t) represents the output result at time t, for example, dynamically changing indicators such as fatigue score and health risk;

[0115]

[0116] This part calculates each factor X i (t), and the influence of historical changes is added. The exponential decay term reflects that the influence of historical changes on the current output gradually decays with time.

[0117] Among them, w i is the weight of factors such as children's physiological characteristics and behavioral habits, i (t) is derived from environmental sensors (such as light, temperature and humidity) and children’s behavioral characteristics (such as eye time, rest time, etc.). λ is the rate constant that controls the reaction of the personalized dynamic early warning model. It is usually a positive number and determines the response rate of the change. F(ΔX(τ)) is a function that reflects changes in fatigue, visual burden or other factors. A memory decay process that reflects changes in features over a period of time in the past;

[0118] in, Combining different features Y j The weighted sum of (t) represents the influence of the feature on the final output;

[0119] It should be noted that the output P(t) of the personalized dynamic early warning model is determined by the weighted sum of the current factors (numerator) and the weighted sum of the feature data (denominator). The historical data (through the exponential decay term) has a gradually decaying effect on the output.

[0120] Use reinforcement learning algorithms to train personalized dynamic warning models;

[0121] It should be noted that by collecting bioelectric signal data (such as electroencephalogram, electrocardiogram, electromyogram) and behavioral characteristics (such as eye time, sitting posture) of children and adolescents in real time, combined with personal health records, a personalized feature vector is constructed, and the bioelectric signal data and behavioral characteristics are input into the reinforcement learning algorithm for training;

[0122] Specifically, reinforcement learning uses a reward function to evaluate the performance of a personalized dynamic warning model by defining state space (such as fatigue scores and environmental conditions) and action space (such as adjusting environmental parameters and sounding an alarm). When the myopia warning signal output by the personalized dynamic warning model is effective, the system will give a positive reward; otherwise, a negative reward will be given. Through multiple training and feedback, the reinforcement learning algorithm continuously adjusts the parameters of the personalized dynamic warning model.

[0123] Based on the training results of the personalized dynamic warning model, the myopia warning threshold is set through a dynamic adjustment strategy.

[0124] Specifically, when the personalized dynamic warning model identifies fatigue signals or other relevant health indicators reaching a certain critical value, it will dynamically adjust the myopia warning threshold to ensure that the warning signal is issued in the most appropriate situation for the individual;

[0125] Preferably, the dynamic adjustment mechanism enables the personalized dynamic early warning model to adapt to the changes and differences of each child and adolescent, avoiding the blindness and inaccuracy of traditional static threshold setting, and improving the accuracy and response speed of the early warning.

[0126] S4. Analyze the current environmental conditions based on the environmental sensor data and the myopia warning threshold, and adjust the environmental conditions in real time through associated smart devices.

[0127] Furthermore, environmental sensors are used to collect data on light, temperature, humidity, and air quality in real time;

[0128] Specifically, a light sensor (such as a photoelectric sensor or a light intensity sensor) is used to monitor the ambient light intensity, which is usually expressed in lux.

[0129] Temperature and humidity sensors (such as DHT11, DHT22) can collect temperature and humidity data in the environment in real time;

[0130] Air quality sensors (such as MQ series sensors and CCS811 sensors) can detect harmful gases (such as carbon dioxide and volatile organic compounds) and particulate matter (such as PM2.5 and PM10) in the air. These sensors output the air quality index (AQI) or pollutant concentration values ​​to indicate the quality of air.

[0131] Analyze current environmental conditions by combining environmental sensor data and myopia warning thresholds;

[0132] It should be noted that the myopia warning threshold refers to a standard value set by a personalized dynamic warning model for the individual health status of children and adolescents. When environmental conditions or behavioral characteristics reach this standard, a myopia warning signal will be triggered.

[0133] The Q-learning algorithm is used to calculate and analyze the results and generate an optimized environmental adjustment strategy;

[0134] It should be noted that the Q-learning algorithm, as a reinforcement learning algorithm, optimizes decision-making strategies through reward and punishment mechanisms, and can continuously adjust its action plans in a dynamic and complex environment to achieve the effect of maximizing the goal. Q-learning generates environmental adjustment strategies related to individual myopia warning thresholds by analyzing environmental data;

[0135] The best is to achieve personalized and dynamic adjustment through the Q-learning algorithm, taking into account the physiological differences and individual needs of children and adolescents under different environmental conditions, not only optimizing the warning threshold setting, but also intelligently adjusting environmental conditions to alleviate the risk of myopia. For example, when the light is strong, it automatically adjusts the indoor lighting to the appropriate brightness, or when the humidity is low, it adjusts the humidity through intelligent humidification equipment to ensure the health of children's eyes;

[0136] Based on the optimized environmental adjustment strategy, by associating intelligent devices, the device parameters are automatically adjusted and the environmental conditions are controlled in real time.

[0137] Specifically, the temperature and humidity sensor monitors the indoor temperature and humidity in real time. For example, if the humidity is lower than 30% or higher than 70%, it may cause dry eyes and fatigue, affecting eye comfort. Set a suitable temperature and humidity range (such as: temperature 22℃ to 26℃, humidity 40% to 60%). When the temperature and humidity exceed the set range, adjust them through smart air conditioning, humidifier or dehumidifier;

[0138] The air quality sensor monitors the indoor air quality index (AQI) in real time, especially the concentration of harmful substances such as PM2.5 and carbon dioxide. If the indoor air quality is poor (for example, the PM2.5 concentration exceeds 50μg / m 3 ), which may cause dry eyes, irritation and fatigue in children and adolescents, and even increase the risk of myopia. According to the air quality monitoring results, if the air quality index exceeds the preset health threshold (for example, AQI>100), the air purifier will be activated and the wind speed will be adjusted to reduce the concentration of harmful gases and particulate matter in the room;

[0139] Preferably, the present invention combines environmental sensor data with a personalized dynamic early warning model and intelligently regulates the environment based on an adjustment strategy generated by a Q-learning algorithm, thereby achieving personalized and adaptive myopia prevention and control. The Q-learning optimization strategy can perform real-time adaptive adjustments in a constantly changing environment, ensuring that every child and adolescent uses their eyes under optimal environmental conditions, thereby effectively reducing the risk of myopia.

[0140] S5. Real-time monitoring of the fatigue scores of children and adolescents. When the fatigue score exceeds the myopia warning threshold, the warning signal is automatically triggered and prompts and vision protection suggestions are generated for the current myopia risk of children and adolescents.

[0141] Furthermore, under environmental condition monitoring, a personalized dynamic early warning model is used to collect and analyze the fatigue scores of children and adolescents in real time;

[0142] Specifically, by wearing wearable devices (such as smart bracelets and smart glasses), children and adolescents can collect electroencephalograms (EEGs), electrocardiograms (ECGs), and electromyograms (EMGs) in real time. Bioelectric signals reflect the physical fatigue state of children and adolescents. For example, within a day, if a child's EEG data shows signs of fatigue in brain wave activity, the fluctuation of his brain wave activity from 10:00 to 12:00 is large and the frequency is low, indicating a high state of fatigue.

[0143] Environmental sensor data showed that the indoor light intensity reached more than 1000 lux at noon, which is strong light and will aggravate eye fatigue;

[0144] Through the personalized dynamic early warning model, the fatigue score is calculated in real time. For example, in the above scenario, assume that the personalized dynamic early warning model calculates the child's fatigue score as 85 (out of 100) based on its EEG and ECG data, and the score exceeds the preset threshold of 70, which means that the child is currently in a state of fatigue and is at risk of visual impairment;

[0145] In a group of 100 children and adolescents, suppose 60% of the children scored above 70 points in a specific period of time (most of these children concentrate on studying for more than 2 hours), and 30% of the children scored above 80 points, showing obvious signs of fatigue;

[0146] According to environmental data analysis, when the light intensity exceeds 800 lux (i.e. when the indoor light source is strong), the average fatigue score increases by 15%. Excessive light increases visual fatigue;

[0147] When the fatigue score exceeds the myopia warning threshold, the warning signal is automatically triggered;

[0148] It should be noted that when the fatigue score obtained by the personalized dynamic warning model analysis exceeds the set myopia warning threshold, a warning signal is automatically triggered to remind the user that the current fatigue state may pose a risk to vision health;

[0149] Preferably, by setting reasonable thresholds and automatically triggering warning signals, it can effectively prevent children and adolescents from continuing to engage in activities that may cause serious visual damage when they are overly tired. Intelligent warnings based on real-time data can more accurately reflect the individual's actual fatigue level, provide timely feedback, and avoid hidden dangers to vision health.

[0150] When the warning signal is triggered, the associated smart device issues a reminder about the current myopia risk to alert children and adolescents;

[0151] Specifically, smart devices push signals and instructions to devices through Bluetooth signals (such as smart watches, smart glasses, and smart bracelets), Wi-Fi signals (such as smart lights, smart air conditioners, and smart speakers), MQTT protocol (message queue telemetry transmission protocol), and HTTP / HTTPS protocol;

[0152] Input myopia risk reminders, fatigue scores, smart devices and bioelectric signal data feedback into a personalized dynamic early warning model to further optimize the early warning strategy;

[0153] It should be noted that all relevant data (such as fatigue scores, smart device data, environmental data, and bioelectric signals) will be fed back to the personalized dynamic early warning model, and the model will dynamically adjust and optimize the early warning strategy based on these data to achieve more efficient monitoring and early warning;

[0154] Preferably, real-time data will be fed back into a personalized dynamic warning model to further optimize the personalized warning strategy so that the warning strategy can adapt to the personalized needs of different users. With the continuous training of the model and the accumulation of data, the accuracy and adaptability of the warning will gradually improve, which will help reduce false alarms and missed reports and improve the vision protection effect of children and adolescents;

[0155] Generate vision protection recommendations based on the optimized early warning strategy and real-time data.

[0156] It should be noted that based on the combination of optimized dynamic warning strategies and real-time data, personalized vision protection suggestions will be automatically generated, such as reasonable eye use time, rest frequency, appropriate sitting posture, light intensity, etc., to help children and adolescents protect their vision scientifically and reasonably;

[0157] Preferably, the steps for generating vision protection recommendations combine all environmental conditions, fatigue scores, bioelectric signal data and other health data to provide comprehensive vision protection measures for children and adolescents; unlike traditional preventive measures, by providing scientific recommendations by comprehensively considering various data, it can effectively reduce the risks of problems such as excessive eye use and myopia development, and constantly make self-adjustments based on actual feedback, making protection measures more precise and personalized, further improving the accuracy and scientific nature of health management.

[0158] The present embodiment also provides a myopia warning system based on children and adolescents, including: a data acquisition module, a feature extraction module, a model building module, an environmental control module and an early warning prompt module; the data acquisition module is used to collect bioelectric signal data and environmental sensor data in real time; the feature extraction module is used to extract features from the collected bioelectric signal data through a combination of a convolutional neural network and a recurrent neural network, identify fatigue signals and fatigue levels, and evaluate the current visual fatigue state of children and adolescents, and output a fatigue score; the model building module is used to build a personalized dynamic early warning model based on the fatigue score and the personal health records of children and adolescents, combined with a reinforcement learning algorithm, according to the physiological characteristics and behavioral characteristics of children in the bioelectric signal data, and set a myopia warning threshold; the environmental control module is used to analyze the current environmental conditions in combination with the environmental sensor data and the myopia warning threshold, and to control the environmental conditions in real time by associating intelligent devices; the early warning prompt module is used to monitor the fatigue score of children and adolescents in real time, and automatically trigger a warning signal when the fatigue score exceeds the myopia warning threshold, and generate prompts and vision protection suggestions for the current myopia risk of children and adolescents.

[0159] This embodiment also provides a computer device, which is suitable for the situation based on the myopia early warning method for children and adolescents, including: a memory and a processor; the memory is used to store computer executable instructions, and the processor is used to execute computer executable instructions to implement the myopia early warning method for children and adolescents proposed in the above embodiment.

[0160] The computer device may be a terminal, and the computer device includes a processor, a memory, a communication interface, a display screen and an input device connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be achieved through WIFI, an operator network, NFC (near field communication) or other technologies. The display screen of the computer device may be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device may be a touch layer covering the display screen, or a key, trackball or touchpad provided on the housing of the computer device, or an external keyboard, touchpad or mouse, etc.

[0161] The present embodiment also provides a storage medium on which a computer program is stored. When the program is executed by a processor, the method for early warning of myopia in children and adolescents proposed in the above embodiment is implemented; the storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (Static Random Access Memory, referred to as SRAM), electrically erasable programmable read-only memory (Electrically Erasable Programmable Read-Only Memory, referred to as EEPROM), erasable programmable read-only memory (Erasable Programmable Read Only Memory, referred to as EPROM), programmable read-only memory (Programmable Red-Only Memory, referred to as PROM), read-only memory (Read-Only Memory, referred to as ROM), magnetic storage, flash memory, disk or optical disk.

[0162] In summary, the present invention proposes a personalized dynamic early warning model by combining bioelectric signals, environmental sensor data and deep learning algorithms, which can monitor the visual fatigue status of children and adolescents in real time and make intelligent adjustments. Compared with the traditional static threshold early warning, the present invention effectively improves the recognition accuracy of fatigue signals through the feature extraction method combining convolutional neural network (CNN) and recurrent neural network (RNN), and can dynamically adapt to the differences in individual physiological characteristics. On this basis, combined with the reinforcement learning algorithm, the early warning model can be intelligently adjusted according to the child’s health record and real-time data, and personalized health intervention measures can be provided. In particular, the combination with environmental data can automatically adjust the environmental conditions in combination with intelligent devices when monitoring fatigue scores in real time, optimize children’s learning and living environment, and further reduce the risk of myopia.

[0163] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A method for early warning of myopia in children and adolescents, characterized by: include, Collect bioelectric signal data and environmental sensor data in real time; By combining convolutional neural networks with recurrent neural networks, the collected bioelectric signal data is feature extracted to identify fatigue signals and fatigue levels, and the current visual fatigue status of children and adolescents is evaluated, and a fatigue score is output; Based on fatigue scores and personal health records of children and adolescents, combined with reinforcement learning algorithms, a personalized dynamic warning model is built according to the physiological and behavioral characteristics of children in bioelectric signal data, and a myopia warning threshold is set; Analyze current environmental conditions by combining environmental sensor data and myopia warning thresholds, and adjust environmental conditions in real time through associated smart devices; Monitor the fatigue scores of children and adolescents in real time. When the fatigue score exceeds the myopia warning threshold, it will automatically trigger a warning signal and generate reminders and vision protection recommendations for the current myopia risk of children and adolescents.

2. The method for early warning of myopia in children and adolescents according to claim 1, characterized in that: The bioelectric signal data includes electroencephalogram (EEG), electrocardiogram (ECG) and electromyogram (EMG); The environmental sensor data includes temperature and humidity sensors, gas sensors and light sensors.

3. The method for early warning of myopia in children and adolescents according to claim 2, characterized in that: The convolutional neural network is combined with the recurrent neural network to extract features from the collected bioelectric signal data and identify fatigue signals and fatigue levels. The specific steps are as follows: Remove noise and standardize bioelectric signal data to retain effective bioelectric signals; Extract the convolutional neural network spatial features from effective bioelectric signals, and use wavelet transform to enhance the high-frequency and low-frequency features of bioelectric signals; Input high-frequency features and low-frequency features into the LSTM network, learn the temporal features of high-frequency features and low-frequency features, then fuse the temporal features through the fully connected layer and output the fused features; The fused features are input into the fully connected network to identify the fatigue signal. SoftMax receives the fatigue signal and outputs the classification probability of the fatigue degree.

4. The method for early warning of myopia in children and adolescents according to claim 3, characterized in that: The specific steps of evaluating the current visual fatigue state of children and adolescents and outputting the fatigue score are as follows: Extract the probability value of each fatigue level from the classification probability output by the fully connected network; Set corresponding scoring intervals of mild fatigue, moderate fatigue and severe fatigue for each fatigue level; Using the classification probability value of each fatigue level, the corresponding scoring intervals are weighted averaged to obtain the final fatigue score.

5. The method for early warning of myopia in children and adolescents according to claim 4, characterized in that: Based on fatigue scores and personal health records of children and adolescents, combined with reinforcement learning algorithms, a personalized dynamic warning model is constructed according to the physiological and behavioral characteristics of children in bioelectric signal data, and a myopia warning threshold is set. The specific steps are as follows: Collect the daily eye use time and sitting posture behavior characteristics of children and adolescents in real time; Combine the bioelectric signal data and behavioral characteristics of children and adolescents to build personal health records; The data in the personal health record is input into the multimodal neural network for feature fusion to obtain a comprehensive personalized feature vector; According to the personalized feature vector, bioelectric signal data and behavioral characteristics, a personalized dynamic early warning model is constructed, which is expressed as: F(ΔX(τ))=ΔX(τ) 2 ; Among them, P(t) is the output of the personalized dynamic warning model at time t, n is the bioelectric signal feature, m is the number of behavioral features, X i (t) is the value of the i-th feature in the bioelectric signal data at time t, w i is the weight corresponding to the i-th feature, Y j (t) is the value of the jth feature in the behavior feature at time t, v j is the weight corresponding to the jth feature, ΔX(τ) is the change in the time interval, τ is the integral variable, dτ is the small change, λ is the decay parameter, and F(ΔX(τ)) is the time difference function; Use reinforcement learning algorithms to train personalized dynamic warning models; Based on the training results of the personalized dynamic warning model, the myopia warning threshold is set through a dynamic adjustment strategy.

6. The method for early warning of myopia in children and adolescents according to claim 5, characterized in that: The specific steps of analyzing the current environmental conditions by combining the environmental sensor data and the myopia warning threshold and regulating the environmental conditions in real time by associating smart devices are as follows: Collect light, temperature, humidity and air quality data in real time through environmental sensors; Analyze current environmental conditions by combining environmental sensor data and myopia warning thresholds; The Q-learning algorithm is used to calculate and analyze the results and generate an optimized environmental adjustment strategy; Based on the optimized environmental adjustment strategy, by associating intelligent devices, the device parameters are automatically adjusted and the environmental conditions are controlled in real time.

7. The method for early warning of myopia in children and adolescents according to claim 6, characterized in that: The real-time monitoring of the fatigue scores of children and adolescents, when the fatigue scores exceed the myopia warning threshold, automatically triggers a warning signal and generates prompts and vision protection suggestions for the current myopia risk of children and adolescents. The specific steps are as follows: Under environmental condition monitoring, a personalized dynamic early warning model is used to collect and analyze children and adolescents’ fatigue scores in real time; When the fatigue score exceeds the myopia warning threshold, the warning signal is automatically triggered; When the warning signal is triggered, the associated smart device issues a reminder about the current myopia risk to alert children and adolescents; Input myopia risk reminders, fatigue scores, smart devices and bioelectric signal data feedback into a personalized dynamic early warning model to further optimize the early warning strategy; Generate vision protection recommendations based on the optimized early warning strategy and real-time data.

8. A myopia early warning system for children and adolescents, based on the myopia early warning method for children and adolescents according to any one of claims 1 to 7, characterized in that: Including data collection module, feature extraction module, model building module, environment control module and early warning module; The data acquisition module is used to collect bioelectric signal data and environmental sensor data in real time; The feature extraction module is used to extract features from the collected bioelectric signal data by combining a convolutional neural network with a recurrent neural network, identify fatigue signals and fatigue levels, and evaluate the current visual fatigue state of children and adolescents, and output a fatigue score; The model building module is used to build a personalized dynamic warning model based on fatigue scores and personal health records of children and adolescents, combined with a reinforcement learning algorithm, according to the physiological and behavioral characteristics of children in bioelectric signal data, and set a myopia warning threshold; The environmental control module is used to analyze the current environmental conditions in combination with the environmental sensor data and the myopia warning threshold, and to control the environmental conditions in real time by associating with the smart device; The early warning prompt module is used to monitor the fatigue scores of children and adolescents in real time. When the fatigue score exceeds the myopia warning threshold, it automatically triggers a warning signal and generates prompts and vision protection suggestions for the current myopia risk of children and adolescents.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method for early warning of myopia in children and adolescents according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method for early warning of myopia in children and adolescents described in any one of claims 1 to 7 are implemented.

Images