Teenager myopia prevention and control intelligent glasses

Abstract

This invention provides a smart glasses for myopia prevention and control in teenagers, relating to the field of smart glasses technology. It includes a nose bridge and a frame. The nose bridge houses a main control module, and an auxiliary camera module is mounted on the top of the nose bridge to capture the wearer's facial expressions. Ambient light sensors are mounted on the top of the front sides of both sides of the frame, and temple mounting heads are mounted on both sides of the frame. Ultraviolet sensors are mounted on the outer sides of the temple mounting heads. Composite lenses are snapped into the inside of the frame. A main control system is embedded within the main control module. This invention utilizes the auxiliary camera module, ambient light sensor, ultraviolet sensor, biosensor, and non-contact eye posture sensor to comprehensively collect multi-dimensional data such as facial expressions, ambient light, ultraviolet intensity, physiological indicators, and eye posture, providing rich and accurate information for subsequent analysis and decision-making.

Description

Technical Field

[0001] This invention relates to the field of smart glasses technology, specifically to a smart glasses for myopia prevention and control in teenagers. Background Technology

[0002] In recent years, myopia among teenagers has shown a serious trend of high incidence and younger age of onset. A research report from the World Health Organization shows that there are currently as many as 600 million people with myopia in China, with the highest rate of myopia among teenagers in the world. This is mainly due to changes in modern lifestyles, which have drastically altered the visual environment for teenagers. With the widespread use of electronic devices such as smartphones and tablets, it is extremely common for children to spend long periods of time using their eyes at close range. Furthermore, academic pressure also leads to prolonged reading and writing, preventing their eyes from getting sufficient rest.

[0003] According to Chinese Patent No. CN 118818791 A, a pair of smart glasses includes a frame, temples, a circuit board, and an elastic element. The frame has a temple connecting portion, which is hollow inside and has a first contact surface. The temples are rotatably connected to the temple connecting portion, which is also hollow inside, and the temples and / or the temple connecting portion form a first channel. The temples have a second contact surface. The circuit board is housed inside the temples, and an electrical connection wire is connected to the circuit board, passing through the first channel into the interior of the temple connecting portion. The elastic element is located outside the first channel and is constrained between the temples and the temple connecting portion, and is in a pre-compressed state so that the second contact surface always tends to approach the first contact surface in the extension direction of the rotation axis of the temples.

[0004] Existing smart glasses have certain intelligent features, but the following problems exist in actual use:

[0005] 1. Existing myopia prevention and control products and devices do not collect comprehensive data, only collecting data from a few aspects, unlike the multi-dimensional data collection provided by the smart glasses in this solution. This results in a lack of in-depth understanding of adolescents' eye use and makes it difficult to formulate precise prevention and control strategies. Even if traditional products collect data, they lack effective analysis and intelligent decision-making mechanisms. They cannot automatically change the lens color, nor can they provide personalized suggestions based on a comprehensive analysis of the wearer's real-time physiological state, eye use behavior, and environmental factors.

[0006] 2. Existing health monitoring and eye posture monitoring devices cannot deeply integrate and analyze data, failing to provide a comprehensive and scientific basis for myopia prevention and control. Furthermore, current myopia prevention and control products are functionally limited, with each device operating independently and lacking collaborative capabilities. Anti-blue light glasses, smart bracelets, and distance sensors on study desks and chairs cannot share data or link functions, making it difficult to create a comprehensive and intelligent eye-protection environment and achieve optimal prevention and control effects.

[0007] Therefore, a smart glasses for myopia prevention and control in teenagers is needed to solve the above problems. Summary of the Invention

[0008] Technical problems to be solved

[0009] To address the shortcomings of existing technologies, this invention provides a smart glasses for myopia prevention and control in teenagers, solving the problems mentioned in the background section.

[0010] Technical solution

[0011] To achieve the above objectives, the present invention provides the following technical solution: a smart glasses for myopia prevention and control in teenagers, comprising a nose bridge and a frame. The nose bridge houses a main control module, and an auxiliary camera module is mounted on the top of the nose bridge for capturing the wearer's facial expressions. Ambient light sensors are mounted on the top of both sides of the frame, and temple mounting heads are mounted on both sides of the frame. Ultraviolet sensors are mounted on the outer sides of the temple mounting heads. Composite lenses are snapped into the frame. A main control system is embedded within the main control module. The main control system includes a data acquisition terminal, a data preprocessing submodule, a health status assessment submodule, an intelligent decision-making submodule, and an execution module. The data acquisition terminal consists of the auxiliary camera module, ambient light sensor, ultraviolet sensor, biosensor, and a newly added non-contact eye posture sensor, which collects the wearer's facial expression data, ambient light data, ultraviolet intensity data, physiological data, and eye posture data in real time.

[0012] Preferably, biosensors are installed at the bottom of the back sides of both sides of the frame, and the biosensors include a photoelectric heart rate sensor, a blood pressure sensor and a blood oxygen saturation sensor.

[0013] Preferably, an electro-optic clip is provided in the middle of the composite lens, and a sandwich structure is formed between the electro-optic clip and the composite lens. A ring-shaped connecting line is provided inside the frame of the lens, and the ring-shaped connecting line is connected to the electro-optic clip. The end of the ring-shaped connecting line is connected to the main control module.

[0014] Preferably, the non-contact eye posture sensor uses infrared sensing and image recognition technology to monitor the distance and angle between the wearer's head and the book or electronic device. The collected data is transmitted to the data preprocessing submodule via a wireless link. The execution module includes a lens color adjustment submodule.

[0015] Preferably, the data preprocessing submodule employs an adaptation algorithm to process data from different sensors, including the auxiliary camera module, ambient light sensor, ultraviolet sensor, biosensor, and the newly added non-contact eye posture sensor. For facial expression data collected by the auxiliary camera module, a convolutional neural network algorithm combining a generative adversarial network (GAN) and an attention mechanism is used for feature extraction and emotion recognition. Adversarial training between the generator and discriminator improves the accuracy of facial expression feature extraction, while the attention mechanism focuses on key facial expression regions. For data from the ambient light sensor and ultraviolet sensor, variational mode decomposition (VMD) combined with a particle filter algorithm is used to remove noise interference. To effectively handle non-Gaussian noise, the system precisely extracts the changing trends of ambient light intensity and ultraviolet radiation, performs adaptive mode decomposition on the signal, and uses particle filtering to effectively process the noise. For data from biosensors, the system utilizes a Long Short-Term Memory (LSTM) network combined with an empirical wavelet transform algorithm for multi-band analysis, separating different physiological signal components. The LSTM network captures the long-term dependencies of physiological signals, and the empirical wavelet transform achieves effective signal decomposition. For data from non-contact eye posture sensors, the system employs a Hough transform-based deep learning target detection algorithm to identify abnormal eye postures. The Hough transform detects lines and angles, while the deep learning target detection algorithm accurately locates the head and device positions.

[0016] Preferably, the health status assessment submodule constructs a composite model comprising a Deep Belief Network (DBN), a Gated Recurrent Unit (GRU), and an Extreme Gradient Boosting (XGBoost) system. This model comprehensively considers eye posture, emotional state, physiological indicators, and environmental factors to assess the wearer's health status. Preprocessed data is input into the model, which then provides a health risk level score based on trained weights and thresholds. When the score falls below a set threshold, an early warning mechanism is immediately triggered. The Deep Belief Network performs unsupervised pre-training of features, the Gated Recurrent Unit processes time-series data, and the Extreme Gradient Boosting system performs efficient classification and regression.

[0017] Preferably, the intelligent decision-making submodule combines a whale optimization algorithm, a simulated annealing algorithm, and a Q-learning algorithm to construct a decision-making model. Through deep learning of historical data, combined with health status assessment results and real-time environmental data, it provides wearers with personalized eye care suggestions, including adjusting screen time and changing the screen environment. The final eye care instructions are then transmitted to the execution module. The whale optimization algorithm performs a global search, the simulated annealing algorithm avoids local optima, and the Q-learning algorithm achieves dynamic optimization of the decision.

[0018] Preferably, the lens color control submodule in the execution module responds immediately upon receiving an instruction, and precisely controls the voltage parameters of the electro-optic clip through the ring connection line inside the lens frame. Under strong light, the voltage of the electro-optic clip is increased to deepen the color of the composite lens, and under dim light, the voltage of the electro-optic clip is decreased to lighten the color of the lens, thereby improving light transmittance and ensuring clear vision for the wearer.

[0019] Preferably, the execution module establishes a communication connection with external smart devices, including a smart desk lamp and a smart study table and chair. When the execution module controls the electro-optic clip to change the lens color, it simultaneously sends this information to the smart desk lamp. Based on the received information and its own light sensor data, the smart desk lamp automatically adjusts the brightness, color temperature, and illumination angle of the light, coordinating with the lens color adjustment of the smart glasses to create a more comfortable and eye-protecting environment.

[0020] Beneficial effects

[0021] This invention provides a smart glasses for myopia prevention and control in teenagers. It has the following beneficial effects:

[0022] 1. This invention utilizes an auxiliary camera module, ambient light sensor, ultraviolet sensor, biosensor, and non-contact eye posture sensor to comprehensively collect multi-dimensional data on facial expressions, ambient light, ultraviolet intensity, physiological indicators, and eye posture, providing rich and accurate information for subsequent analysis and decision-making. For different sensor data, it employs appropriate advanced algorithms, such as GAN combined with attention mechanisms in convolutional neural networks to process facial expression data, and VMD combined with particle filtering to process ambient light and ultraviolet data, significantly improving data quality and usability, laying a solid foundation for assessing health status and developing eye care strategies.

[0023] 2. The lens color adjustment submodule of the execution module of this invention, based on the instructions of the intelligent decision-making submodule, precisely controls the electro-clamp voltage parameters in real time to achieve rapid adjustment of the composite lens color. Under strong light, the lens color deepens to block excessive light and protect the eyes; in dim light, the lens color lightens to improve light transmittance. Furthermore, based on health status assessment results, the lens color can be adjusted to a specific hue to relieve eye fatigue, providing comprehensive eye care.

[0024] 3. The intelligent decision-making submodule of this invention combines whale optimization, simulated annealing, and Q-learning algorithms. By deeply learning historical data and analyzing real-time environmental data, it provides wearers with personalized eye care suggestions. This intelligent decision-making mechanism can formulate appropriate strategies for adjusting screen time and changing the screen environment based on different scenarios and individual differences, effectively preventing the occurrence and development of myopia. Attached Figure Description

[0025] Figure 1 This is a system framework diagram of the main control module of the present invention;

[0026] Figure 2 This is an overall structural diagram of the present invention;

[0027] Figure 3 This is an overall rear view of the present invention;

[0028] Figure 4 This is a structural diagram of the components of the present invention;

[0029] Figure 5 This is the control logic diagram of the execution module of the present invention.

[0030] Legend:

[0031] 1. Nose bridge; 2. Eyeglass frame; 3. Ambient light sensor; 4. Temple mount; 5. Ultraviolet sensor; 6. Composite lens; 7. Biosensor; 8. Auxiliary camera module; 9. Electro-optic clip; 10. Ring connection circuit. Detailed Implementation

[0032] 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. Specific Implementation Example 1:

[0034] like Figure 1-5As shown, a smart glasses for myopia prevention and control in teenagers includes a nose bridge 1 and a frame 2. The nose bridge 1 has a main control module inside, and an auxiliary camera module 8 is installed on the top of the nose bridge 1. The auxiliary camera module 8 is used to capture the wearer's facial expressions. Ambient light sensors 3 are installed on the top of the front of both sides of the frame 2. Temple mounting heads 4 are installed on both sides of the frame 2. Ultraviolet sensors 5 are installed on the outer side of the temple mounting heads 4. Composite lenses 6 are snapped into the inside of the frame 2. The main control module has a main control system embedded in it. The main control system includes a data acquisition end, a data preprocessing submodule, a health status assessment submodule, an intelligent decision-making submodule, and an execution module. The data acquisition end consists of the auxiliary camera module 8, ambient light sensor 3, ultraviolet sensor 5, biosensor 7, and a newly added non-contact eye posture sensor, which collects the wearer's facial expression data, ambient light data, ultraviolet intensity data, physiological data, and eye posture data in real time.

[0035] Biosensors 7 are installed at the bottom of the back on both sides of the frame 2. The biosensors 7 include a photoelectric heart rate sensor, a blood pressure sensor, and a blood oxygen saturation sensor.

[0036] An electro-optic clip 9 is provided in the middle of the composite lens 6, forming a sandwich structure between the electro-optic clip 9 and the composite lens 6. A ring-shaped connecting line 10 is provided inside the frame of the frame 2, which is connected to the electro-optic clip 9. The end of the ring-shaped connecting line 10 is connected to the main control module.

[0037] The non-contact eye posture sensor uses infrared sensing and image recognition technology to monitor the distance and angle between the wearer's head and the book or electronic device. The collected data is transmitted to the data preprocessing submodule via a wireless link, and the execution module includes a lens color adjustment submodule.

[0038] The data preprocessing submodule employs adaptive algorithms to process data from various sensors, including the auxiliary camera module 8, ambient light sensor 3, ultraviolet sensor 5, biosensor 7, and the newly added non-contact eye posture sensor. For facial expression data collected by the auxiliary camera module 8, a convolutional neural network algorithm combining a generative adversarial network (GAN) and an attention mechanism is used for feature extraction and emotion recognition. Adversarial training between the generator and discriminator improves the accuracy of facial expression feature extraction, while the attention mechanism focuses on key facial expression regions. For data from the ambient light sensor 3 and ultraviolet sensor 5, variational mode decomposition (VMD) combined with a particle filter algorithm is used to remove noise interference. To eliminate noise, the system accurately extracts the changing trends of ambient light intensity and ultraviolet radiation, performs adaptive mode decomposition on the signal, and effectively handles non-Gaussian noise using particle filtering. For data from the biosensor 7, it utilizes a Long Short-Term Memory (LSTM) network combined with an empirical wavelet transform algorithm for multi-band analysis, separating different physiological signal components. The LSTM network captures the long-term dependencies of physiological signals, and the empirical wavelet transform achieves effective signal decomposition. For data from a non-contact eye posture sensor, it employs a Hough transform-based deep learning target detection algorithm to identify abnormal eye postures. The Hough transform detects lines and angles, and the deep learning target detection algorithm accurately locates the head and device positions.

[0039] The health status assessment submodule constructs a composite model comprising a Deep Belief Network (DBN), a Gated Recurrent Unit (GRU), and Extreme Gradient Boosting (XGBoost). This model comprehensively considers eye posture, emotional state, physiological indicators, and environmental factors to assess the wearer's health status. Preprocessed data is input into the model, which then provides a health risk level score based on the trained weights and thresholds. When the score falls below a set threshold, an early warning mechanism is immediately triggered. The Deep Belief Network performs unsupervised pre-training of features, the Gated Recurrent Unit processes time-series data, and Extreme Gradient Boosting performs efficient classification and regression.

[0040] The intelligent decision-making submodule combines the whale optimization algorithm, simulated annealing algorithm, and Q-learning algorithm to construct a decision-making model. Through deep learning of historical data, combined with health status assessment results and real-time environmental data, it provides wearers with personalized eye care suggestions, including adjusting screen time and changing the screen environment. The final eye care instructions are then transmitted to the execution module. The whale optimization algorithm performs a global search, the simulated annealing algorithm avoids local optima, and the Q-learning algorithm achieves dynamic optimization of the decision.

[0041] The lens color control submodule in the execution module responds immediately upon receiving the instruction. Through the ring connection line 10 inside the frame 2, it precisely controls the voltage parameters of the electro-clipping clip 9. Under strong light, the voltage of the electro-clipping clip 9 is increased to deepen the color of the composite lens 6. Under dim light, the voltage of the electro-clipping clip 9 is decreased to lighten the lens color, improve light transmittance, and ensure clear vision for the wearer.

[0042] The execution module establishes a communication connection with external smart devices, including a smart desk lamp and a smart study table and chair. When the execution module controls the electro-optic clip 9 to change the lens color, it simultaneously sends this information to the smart desk lamp. Based on the received information and its own light sensor data, the smart desk lamp automatically adjusts the brightness, color temperature, and illumination angle of the light, coordinating with the lens color adjustment of the smart glasses to create a more comfortable and eye-protecting environment. Specific Implementation Example 2:

[0044] like Figure 1-5 As shown, the following is a supplement to the above embodiments:

[0045] Overall working principle

[0046] Smart glasses continuously collect multi-dimensional data on the wearer's physiological state, visual environment, and visual behavior through various sensors distributed in key locations. The data collected includes facial expression data captured by the auxiliary camera module 8, ambient light data detected by the ambient light sensor 3, ultraviolet intensity data acquired by the ultraviolet sensor 5, physiological data such as heart rate, blood pressure, and blood oxygen saturation measured by the biosensor 7, and distance and angle data between the head and books or electronic devices monitored by a non-contact eye posture sensor.

[0047] These raw data are transmitted to the data preprocessing submodule, where specialized algorithms are used to process different types of sensor data. For example, for facial expression data, a convolutional neural network algorithm combining a generative adversarial network (GAN) with an attention mechanism is used to accurately extract emotional features; ambient light and ultraviolet data are processed by variational mode decomposition (VMD) combined with a particle filter algorithm to remove noise interference and obtain accurate trends in light intensity and ultraviolet intensity; biosensor data is analyzed using a long short-term memory network (LSTM) combined with an empirical wavelet transform algorithm to separate different physiological signal components through multi-band analysis; and non-contact eye posture sensor data is used to identify abnormal eye postures using a target detection algorithm based on Hough transform combined with deep learning.

[0048] The preprocessed data is input into the health status assessment submodule. This module constructs a composite model that includes a deep belief network (DBN), a gated recurrent unit (GRU), and an extreme gradient boosting (XGBoost) system. It comprehensively considers eye posture, emotional state, physiological indicators, and environmental factors to conduct a comprehensive assessment of the wearer's health status and output a health risk level score.

[0049] The intelligent decision-making submodule generates personalized eye care suggestions based on health status assessment results, historical data, and real-time environmental data, using an intelligent decision-making algorithm that combines whale optimization algorithm, simulated annealing algorithm, and Q-learning algorithm, and sends the eye care instructions to the execution module.

[0050] Execution Module and Lens Color Control: After receiving a command, the execution module's lens color adjustment submodule responds quickly. The composite lens 6 adopts a special sandwich structure, with an electro-optic clip 9 in the middle. The electro-optic clip 9 is connected to the composite lens 6 via a ring-shaped connecting line 10 within the frame 2. The end of the ring-shaped connecting line 10 is connected to the main control module. When the lens color adjustment submodule receives a command from the intelligent decision-making submodule, it precisely controls the voltage parameters of the electro-optic clip 9 based on different ambient light conditions and health status assessment results.

[0051] The photochromic principle under strong light: When the ambient light sensor 3 detects excessive light, or the health assessment submodule determines that the current light conditions may cause damage to the eyes, the intelligent decision-making submodule sends a command to the execution module to darken the lens color. The lens color control submodule increases the voltage of the electrochromic clip 9 through the ring connection line 10. The electrochromic clip 9 typically uses an electrochromic material, such as tungsten oxide (WO3). Under the action of an electric field, tungsten oxide undergoes a redox reaction, changing the distribution of ions and electrons within it, thereby altering its light absorption and reflection characteristics. As the voltage increases, more ions are embedded in the tungsten oxide lattice, enhancing its absorption of visible light and reducing reflection. The color of the composite lens 6 darkens accordingly, effectively blocking excessive light from entering the eyes and reducing the stimulation of the retina by strong light.

[0052] The principle of color change in dim lighting: When the ambient light sensor 3 detects dim lighting conditions that do not meet normal eye use requirements, the intelligent decision-making submodule sends a command to the execution module to lighten the lens color. The lens color adjustment submodule reduces the voltage of the electro-optic clip 9 through the ring connection line 10. At this time, ions in the electro-optic clip are released from the crystal lattice, and the ion and electron distribution of tungsten oxide returns to near its initial state. This reduces light absorption, increases reflection, and lightens the color of the composite lens 6, improving light transmittance and ensuring that the wearer can see clearly even in dimly lit environments.

[0053] Color adjustment based on health status: In addition to changes in ambient light, the results of the health status assessment submodule also affect the adjustment of lens color. For example, when the biosensor 7 detects abnormal physiological indicators related to eye fatigue in the wearer, or when the non-contact eye posture sensor detects prolonged poor eye use behavior, even if the ambient light intensity is within the normal range, the intelligent decision-making submodule may instruct the lens color control submodule to adjust the voltage of the electro-clamp 9 based on the comprehensive assessment results, adjusting the lens color to a specific hue that helps relieve eye fatigue, such as a light green tone, to reduce visual fatigue and protect vision.

[0054] Meanwhile, the execution module also establishes a communication connection with external smart devices (smart desk lamp and smart study table and chair). When the lens color changes, the information is sent to the smart desk lamp in a synchronized manner. Based on the received information and its own light sensor data, the smart desk lamp automatically adjusts the brightness, color temperature and illumination angle of the light, working together with the smart glasses to create an eye-protecting environment.

[0055] 2. Additional Sensor Details

[0056] Auxiliary camera module 8: Utilizes a Sony IMX477 image sensor with 12 megapixels, providing high-resolution images to ensure clear capture of the wearer's facial expression details. It is paired with a 2.8mm wide-angle lens, offering approximately 80° of field of view, sufficient to cover the wearer's entire face. When the camera is in operation, it captures images at 30 frames per second and transmits the image data to the main control module via an SPI (Serial Peripheral Interface) interface.

[0057] Ambient Light Sensor 3: Employs the AMSAS7341 ambient light sensor, which can detect a wide spectral range from visible to near-infrared light, featuring high sensitivity and fast response. It can accurately measure ambient light intensity (range 0-65535 lux) and color temperature (range 2700K-6500K). The sensor communicates with the main control module via an I2C (Inter-Integrated Circuit) interface, sending the latest ambient light data to the main control module every 100ms.

[0058] Ultraviolet Sensor 5: The Vishay VCNL4010 ultraviolet sensor was selected. This sensor can accurately measure ultraviolet intensity and has extremely low sensitivity to visible and infrared light. Its measurement range is 0-150 mW / cm². 2 The accuracy can reach ±10%. It connects to the main control module via an I2C interface and updates the ultraviolet intensity data every 200ms.

[0059] Bio-collection sensor 7:

[0060] Photoelectric heart rate sensor: Utilizing the MAX30102 model, it employs the photoelectric volumetric recording (PPG) principle, measuring heart rate by emitting infrared and red light and detecting changes in reflected light. This sensor can quickly and accurately measure heart rate within 10 seconds, with a measurement range of 30-250 BPM and an accuracy of ±2 BPM. It communicates with the main control module via an SPI interface.

[0061] Blood pressure sensor: Utilizing the same oscillometric blood pressure measurement chip as the Omron HEM-7130, it calculates blood pressure by detecting pulse waves caused by pressure changes within the cuff (a similar air pocket structure integrated into the temples of eyeglasses). The measurement range is systolic pressure 60-260 mmHg and diastolic pressure 40-150 mmHg, with a measurement accuracy of ±3 mmHg. Measurement results are transmitted to the main control module via a UART (Universal Asynchronous Receiver / Transmitter) interface, with blood pressure measurements taken every 5 minutes.

[0062] Blood oxygen saturation sensor: Utilizing the Maxim Integrated MAX30105 model, it employs a dual-wavelength light absorption principle to calculate blood oxygen saturation by detecting the absorption of different wavelengths of light in the blood. The measurement range is 70%-100%, with an accuracy of ±2%. It communicates with the main control module via an SPI interface to transmit blood oxygen saturation data in real time.

[0063] Non-contact eye posture sensor: This sensor employs an infrared sensing module consisting of an infrared emitter and receiver, paired with an image recognition processing unit based on the Rockchip RK3399 chip. The infrared emitter emits infrared light of a specific frequency. When the wearer's head approaches a book or electronic device, the infrared light is reflected back. The infrared receiver receives the reflected light and transmits the signal to the image recognition processing unit. This unit uses a Hough transform-based deep learning object detection algorithm to analyze and process the image, identifying the distance and angle between the head and the book or electronic device. The processed eye posture data is transmitted to the main control module via a USB interface, with data updates every 500ms.

[0064] Communication details

[0065] Communication between smart glasses and smartphones: The smart glasses connect to the smartphone via Bluetooth 5.0. The Bluetooth module uses the Nordic RF52832 chip, which boasts low power consumption and a transmission range of up to 100 meters. The smart glasses transmit health status assessment results, eye care suggestions, and real-time sensor data to the smartphone via Bluetooth. Users can view detailed data, set reminders, and adjust smart glasses parameters through a mobile app.

[0066] Communication between smart glasses and external smart devices: The smart glasses communicate with the smart desk lamp and smart study chair via Wi-Fi. The built-in Wi-Fi module of the smart glasses uses a MediaTek MT7601U chip, supporting the 2.4GHz band. When the execution module controls the electro-optic clip 9 to change the lens color, the relevant information is sent to the controllers of the smart desk lamp and smart study chair via Wi-Fi, enabling collaborative work between the devices.

[0067] Power Management Details

[0068] The smart glasses are powered by a built-in 300mAh lithium polymer battery with a voltage of 3.7V. The power management chip used is the Texas Instruments TPS65132, which provides functions such as battery charging management, overvoltage protection, overcurrent protection, and low battery detection. When the battery level drops below 20%, the smart glasses send a low battery alert to the smartphone via Bluetooth and simultaneously reduce the sampling frequency of some sensors to extend battery life. Charging takes approximately 1.5 hours using a 5V / 1A USB charger.

[0069] Software algorithm implementation details

[0070] The data preprocessing submodule includes a convolutional neural network algorithm combining Generative Adversarial Network (GAN) and an attention mechanism. Both the generator and discriminator are constructed using convolutional layers, batch normalization layers, and ReLU activation functions. The attention mechanism focuses attention on key facial expression regions such as the eyes and mouth by calculating attention weights, thereby improving the accuracy of expression feature extraction. In the Variational Mode Decomposition (VMD) algorithm, ambient light and ultraviolet data are decomposed into multiple modal components by setting appropriate penalty parameters and iteration counts. A particle filter algorithm is then used to filter each modal component to remove noise interference. In the Long Short-Term Memory (LSTM) network combined with Empirical Wavelet Transform (EVT) algorithm, the LSTM network has three hidden layers, each containing 128 neurons. The EPT adaptively selects the wavelet basis based on the signal's frequency characteristics, enabling multi-band analysis of biosensor data. In the object detection algorithm based on Hough transform combined with deep learning, Hough transform is used to detect lines and angles in images, while the deep learning object detection algorithm adopts the SSD (SingleShotMultiBoxDetector) model based on convolutional neural networks, which identifies the positions of heads, books, and electronic devices by training the model.

[0071] The health status assessment submodule consists of a Deep Belief Network (DBN) composed of three layers of Restricted Boltzmann Machines (RBMs) that extract features from preprocessed data through unsupervised learning. A Gated Recurrent Unit (GRU) with two layers, each containing 64 neurons, processes time-series data, capturing trends in eye posture, physiological indicators, and other parameters over time. The Extreme Gradient Boosting (XGBoost) model, by setting appropriate parameters such as learning rate, tree depth, and subsample ratio, efficiently classifies and regresses the input data to assess the wearer's health risk level.

[0072] The intelligent decision-making submodule, specifically the whale optimization algorithm, simulates whale predation behavior to perform a global search of the decision space, seeking the optimal eye protection strategy. The simulated annealing algorithm, during the search process, accepts inferior solutions with a certain probability to avoid getting trapped in local optima. The Q-learning algorithm continuously interacts with the environment to learn the optimal decision-making strategy, generating personalized eye protection suggestions based on different health states and environmental conditions. Specific Implementation Example 3:

[0074] like Figure 1-5 As shown, the following are the core mathematical formulas of each algorithm in this scheme, along with their corresponding explanations and descriptions:

[0075] 1. Generative Adversarial Networks (GANs):

[0076] The goal of generator G is to generate data that is as realistic as possible to deceive discriminator D. Its optimization objective is:

[0077]

[0078] Where x is the actual data, p data (x) is the true data distribution, z is random noise, and p z (z) is the noise distribution, G(z) is the data generated by the generator, D(x) is the probability of the discriminator judging the real data, and D(G(z)) is the probability of the discriminator judging the generated data.

[0079] 2. Attention mechanism:

[0080] Assuming the input feature map is F and the attention weight map is A, then the output O of the attention mechanism is:

[0081] O = F × A

[0082] The calculation of attention weights A is typically based on operations such as convolution, for example:

[0083] A = σ(W2·ReLU(W1·F))

[0084] Where W1 and W2 are learnable weight matrices, σ is an activation function (such as sigmoid), and ReLU is a linear rectified function.

[0085] 3. Variational Mode Decomposition (VMD):

[0086] For a given signal f(t), VMD decomposes it into a series of mode functions u with finite bandwidth. k (t), whose constrained variational problem is:

[0087]

[0088] The constraints are Where K is the number of modes, ω k is the center frequency of the k-th mode, δ(t) is the Dirac function, and * denotes convolution.

[0089] 4. Particle filtering:

[0090] Importance sampling: from the importance distribution q(x) t |x 0:t-1 ,z 1:t Sampling in) The weight update formula is:

[0091]

[0092] in z is the weight of the i-th particle at time t. t These are observed values. It is the observational likelihood. It is the state transition probability. It is an importance distribution.

[0093] Resampling: based on weights Resampling the particles yields a new set of particles.

[0094] 5. Whale Optimization Algorithm:

[0095] Surround the prey:

[0096]

[0097] in This is the current optimal solution. This is the current solution. and It is a coefficient vector, and t is the current iteration number.

[0098] Searching for prey:

[0099]

[0100] in It is a randomly selected solution.

[0101] Attacking prey:

[0102]

[0103] in It is a random vector between [-1, 1].

[0104] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising a reference structure" does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes the element.

[0105] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A smart glasses for myopia prevention and control in teenagers, comprising a nose bridge (1) and a frame (2), characterized in that: The nose bridge (1) is equipped with a main control module inside. An auxiliary camera module (8) is installed on the top of the nose bridge (1). An ambient light sensor (3) is installed on the top of the front of both sides of the frame (2). A temple mounting head (4) is installed on both sides of the frame (2). An ultraviolet sensor (5) is installed on the outside of the temple mounting head (4). A composite lens (6) is snapped into the inside of the frame (2). A main control system is embedded in the main control module. The main control system includes a data acquisition end, a data preprocessing submodule, a health status assessment submodule, an intelligent decision-making submodule, and an execution module. The data acquisition end consists of an auxiliary camera module (8), an ambient light sensor (3), an ultraviolet sensor (5), a bio-aggregate sensor (7), and a newly added non-contact eye posture sensor. It collects the wearer's facial expression data, ambient light data, ultraviolet intensity data, physiological data, and eye posture data in real time. The data preprocessing submodule employs an adaptation algorithm to process data from different sensors, including the auxiliary camera module (8), ambient light sensor (3), ultraviolet sensor (5), biosensor (7), and the newly added non-contact eye posture sensor. For the facial expression data collected by the auxiliary camera module (8), a convolutional neural network algorithm combining generative adversarial network (GAN) and attention mechanism is used for feature extraction and emotion recognition. Adversarial training between the generator and discriminator is used to improve the accuracy of facial expression feature extraction. At the same time, the attention mechanism focuses on key facial expression regions. For the data from the ambient light sensor (3) and ultraviolet sensor (5), variational mode decomposition (VMD) combined with particle filtering algorithm is used. The method removes noise interference, accurately extracts the changing trends of ambient light intensity and ultraviolet intensity, performs adaptive mode decomposition on the signal, and particle filtering effectively processes non-Gaussian noise; for the data of the bio-aggregate sensor (7), the Long Short-Term Memory Network (LSTM) combined with the empirical wavelet transform algorithm is used to perform multi-band analysis, separate different physiological signal components, the LSTM captures the long-term dependence of physiological signals, and the empirical wavelet transform achieves effective decomposition of signals; for the data of the non-contact eye posture sensor, the Hough transform combined with the deep learning target detection algorithm is used to identify abnormal eye postures, the Hough transform detects straight lines and angles, and the deep learning target detection algorithm accurately locates the head and device positions.

2. The smart glasses for myopia prevention and control in teenagers according to claim 1, characterized in that: Biosensors (7) are installed at the bottom of the back side of both sides of the frame (2). The biosensors (7) include a photoelectric heart rate sensor, a blood pressure sensor and a blood oxygen saturation sensor.

3. The smart glasses for myopia prevention and control in teenagers according to claim 2, characterized in that: An electro-optic clip (9) is provided in the middle of the composite lens (6), and a sandwich structure is formed between the electro-optic clip (9) and the composite lens (6). A ring-shaped connecting line (10) is provided inside the frame of the lens frame (2), and the ring-shaped connecting line (10) is connected to the electro-optic clip (9). The end of the ring-shaped connecting line (10) is connected to the main control module.

4. The smart glasses for myopia prevention and control in teenagers according to claim 3, characterized in that: The non-contact eye posture sensor uses infrared sensing and image recognition technology to monitor the distance and angle between the wearer's head and the book or electronic device. The collected data is transmitted to the data preprocessing submodule via a wireless link. The execution module includes a lens color adjustment submodule.

5. The smart glasses for myopia prevention and control in teenagers according to claim 4, characterized in that: The health status assessment submodule constructs a composite model comprising a deep belief network (DBN), a gated recurrent unit (GRU), and extreme gradient boosting (XGBoost). This model comprehensively considers eye posture, emotional state, physiological indicators, and environmental factors to assess the wearer's health status. Preprocessed data is input into the model, which then assigns a health risk level score based on trained weights and thresholds. When the score falls below a set threshold, an early warning mechanism is immediately triggered. The deep belief network performs unsupervised pre-training of features, the gated recurrent unit processes time-series data, and extreme gradient boosting performs efficient classification and regression.

6. The smart glasses for myopia prevention and control in teenagers according to claim 5, characterized in that: The intelligent decision-making submodule combines the whale optimization algorithm, simulated annealing algorithm, and Q-learning algorithm to construct a decision-making model. Through deep learning of historical data, combined with health status assessment results and real-time environmental data, it provides wearers with personalized eye care suggestions, including adjusting screen time and changing the screen environment. The final eye care instructions are then transmitted to the execution module. The whale optimization algorithm performs a global search, the simulated annealing algorithm avoids local optima, and the Q-learning algorithm achieves dynamic optimization of the decision.

7. The smart glasses for myopia prevention and control in teenagers according to claim 6, characterized in that: The lens color control submodule in the execution module responds immediately upon receiving the instruction. Through the ring connection line (10) in the frame (2), it precisely controls the voltage parameters of the electro-clamp (9). Under strong light, it increases the voltage of the electro-clamp (9) to deepen the color of the composite lens (6). Under dim light, it decreases the voltage of the electro-clamp (9) to lighten the lens color, improve the light transmittance, and ensure that the wearer can see clearly.

8. The smart glasses for myopia prevention and control in teenagers according to claim 7, characterized in that: The execution module establishes a communication connection with external smart devices, including smart desk lamps and smart study tables and chairs. When the execution module controls the electro-optic clip (9) to change the lens color, it simultaneously sends this information to the smart desk lamp. Based on the received information and combined with its own light sensor data, the smart desk lamp automatically adjusts the brightness, color temperature and illumination angle of the light, which works in conjunction with the lens color adjustment of the smart glasses to create a more comfortable and eye-protecting environment.

Images